Geometry

with
Computers
Computer-Based Techniques to
Learn and Teach
Euclidean Geometry

Tom Davis

Draft Date: May 17, 2006

Chapter

Preface
Mathematics must be written into the mind, not read into it. No head
for mathematics nearly always means Will not use a pencil.
Arthur Latham Baker

0.1 Why Another Geometry Book?

Euclidean Geometry is ancient, and thousands of books and articles have been written
on the subject. Why write another?
Here are the differences between this book and the others:
Most mathematics books are written to communicate results: theorems, calculation methods and algorithms. This book concentrates on computer-based techniques for solving geometric problems.
i

ii

PREFACE
Although it may not be obvious, mathematics is highly geometricalvirtually
every formula has an associated picture. Often it is much easier to obtain understanding by mentally manipulating the picture than by manipulating a formula
algebraically. What better way to practice mental manipulation of pictures than
to study geometry?
Mental manipulation of geometry is difficult for many people, but today almost
any relatively new personal computer has graphical capabilities that would have
been unimaginably powerful 20 years ago. These machines can help visualize
geometric results in a dynamic manner that is far more compelling than the fixed
drawings appearing in standard mathematics books or that can be produced with
pencil and paper. This book comes with a computer program called Geometer
that allows you to manipulate existing geometric diagrams and to create your
own.
Geometer can also be used as an experimental tool to run geometry experiments that can aid greatly in both understanding and proving results.
Although Euclidean geometry is a huge field, it is not a big part of the general
mathematics curriculum in the United States. I have a Ph.D. in mathematics,
and the only formal class I had in Euclidean geometry was in high school. Your
geometry class may have been different, but in mine we somehow managed to
avoid learning most of the truly beautiful results.
High school geometry is often the first introduction students have to constructing
mathematical proofs. It is obviously better to begin with simple examples when
teaching a student to write proofs, so the proof construction exercises in typical
high school texts can almost all be completed in three or four steps.
Unfortunately, if one is limited to proofs of only a few steps, a huge proportion of
geometry is inaccessible. For people who can follow a ten or twenty step proof,
it is amazing what results are accessiblesome of the most beautiful theorems
in mathematics.
Ones technique of geometric proof can be honed with artificial drills, or one can
work on interesting problems. This is similar to the difference between running
on a treadmill or on mountain trailsthe treadmill is completely predictable
while the trails can be tricky but the trail runner will not be bored out of his
mind. This book is aimed at those who did enough treadmill geometry in high
school and want to see some beauty while they learn advanced techniques.
The material here is accessible to brighter high school students, and can be used
as supplementary material by teachers or by the students themselves.
The Geometer program can also be used as a presentation device in geometry
classes. Although most of the material is beyond what is taught in high school,
many of the more important introductory results are available on the enclosed
CD in Geometer format.
There is additional material on the CD to helpsome more elementary theorems, and a tutorial on the construction of sophisticated Geometer diagrams, for
example. See the index file on the CD for up to date information.

0.2. AUDIENCE

iii

Finally, with computer graphics increasing in power every year, Euclidean geometry (and projective geometry as well) may be destined for a comeback. More
and more computer applications are being designed to help people visualize
problems geometrically, but to make the computers do a good job of that, more
knowledge of geometry is required.

0.2 Audience
This book is for teachers of Euclidean geometry but it contains topics of interest for
motivated students or anyone who loves geometry. Students and coaches of students
who compete in high level mathematics competitions should find the advanced material
in this text valuable as training aids.

0.3 Origins of Geometer

The I worked at Silicon Graphics for 16 years doing graphics programming and helping
design computer graphics hardware and so I have a fairly solid grounding in both the
theoretical and practical details of geometry.
Although I was paid money to be a computer engineer, I have never ceased to be a
mathematician as well. From time to time both in and out of work I ran into interesting
geometric theorems. A couple of them seemed so surprising and non-intuitive that I
wrote computer programs to help me visualize them. After going through that exercise
a few times, I realized that it would not be hard to write a general-purpose program
where arbitrary geometric diagrams could be entered with a graphical user interface
(GUI) and dynamically altered. I hacked together the original version of Geometer on
Silicon Graphics UNIX machines in my spare time. (I was wrong when I thought it
would not be hard.)
The version of Geometer included with this book is vastly modified and improved
(and it now runs on Windows, Macintosh OS X and Linux machines), but much of the
underlying philosophy is the same.
As I worked on this book, I tried to practice what I preach. I knew of many interesting results that I had never actually proved, and rather than just look up the standard
proofs, I struggled to figure them out myself. Of course from time to time I had to look
in various books for hints and even when I did solve a problem completely without
hints, I later read the literature to see what the standard methods are.
In my struggles to find the proofs, I used Geometer extensively in exactly the same
ways I expect you the reader to use it. It may not be bug-free, but I can assure you
that it is quite robust, and since it has been used to create thousands of real dynamic
diagrams, it is fairly well streamlined: when I got annoyed with something, I just fixed
it.

PREFACE

iv

0.4 Acknowledgements
A mathematician is a machine for turning coffee into theorems.
Alfred Renyi

Zvezdelina Stankova showed me many wonderful geometric ideas. In particular, I

learned from her the power of the technique of inversion in a circle.
I would like to thank Joshua Zucker and his students in Palo Altos Gunn High
School Math Circle for all the stimulating discussions of mathematics problems and
for serving as unwitting guinea pigs for some of the presentation techniques used here.
I also learned a lot in Zvezdelinas UC Berkeley Math Circle, and in Tatiana Shubins
Math Circle at San Jose State University.
Special thanks go to Gerald Alexanderson and Arthur Benjamin for their help
in converting what began as a disorganized collection of half-baked ideas into the
completely-baked version you hold in your hands. Rudolf Fritsch provided many detailed corrections and suggestions.
I received many wonderful suggestions for improvements and enhancements to the
Geometer program from Paul Haeberli and Henry Moreton.
This book was produced using LATEX, the Emacs editor, and the PostScript files
produced by Geometer. After an infuriating attempt to use the most popular word
processor in the world for the text, Emacs and LATEX were a total pleasure to use, so I
must thank Donald Knuth for writing TEX and Leslie Lamport for writing LATEX.
My wife, Ellyn Bush, although not a mathematician, read and re-read many parts
of the text, and provided many valuable insights on how to make things simpler and
clearer. She put up with all of my complaints with only a few of her own.
Finally, as the quotation at the beginning of this section indicates, I would like to
thank whoever it was who discovered that beans of the shrub Coffea arabica could be
roasted, ground and steeped in boiling water to produce the nectar of life. I converted
that nectar not only into theorems, but also into text and into the computer code that
produced Geometer.
Tom Davis
May, 2004

1 This

result was sharpened by Pal Turan who added that weak coffee produces only lemmas.

Contents

Preface

0.1

Why Another Geometry Book? . . . . . . . . . . . . . . . . . . . . .

0.2

Audience . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

iii

0.3

Origins of Geometer . . . . . . . . . . . . . . . . . . . . . . . . . .

iii

0.4

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . .

iv

1 Introduction

1.1

Computer Geometry Programs . . . . . . . . . . . . . . . . . . . . .

1.2

The Contents of the CD . . . . . . . . . . . . . . . . . . . . . . . . .

1.3

Organization of the Book . . . . . . . . . . . . . . . . . . . . . . . .

1.4

Proofs in this Book . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.5

Illustrations in this Book . . . . . . . . . . . . . . . . . . . . . . . .

1.6

How to Use this Book . . . . . . . . . . . . . . . . . . . . . . . . . .

1.7

Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

1.8

Where to Go from Here . . . . . . . . . . . . . . . . . . . . . . . . .

11

2 Computer-Assisted Geometry

15

2.1

Accurate Drawings . . . . . . . . . . . . . . . . . . . . . . . . . . .

16

2.2

Drawing Manipulation . . . . . . . . . . . . . . . . . . . . . . . . .

21

2.3

Finding and Testing Conjectures . . . . . . . . . . . . . . . . . . . .

22

2.4

conjecture testing . . . . . . . . . . . . . . . . . . . . . . . . . . . .

22

2.5

Ease of Correction . . . . . . . . . . . . . . . . . . . . . . . . . . .

24

2.6

Stepping through Proofs . . . . . . . . . . . . . . . . . . . . . . . .

24

2.7

Making Measurements . . . . . . . . . . . . . . . . . . . . . . . . .

26

CONTENTS

vi
2.8
2.9
2.10
2.11
2.12
2.13
2.14
2.15

Searching for Patterns . . . . . . . . .

Searching for Extremes . . . . . . . .
Locus of Points . . . . . . . . . . . .
Computer-Generated Diagrams . . . .
Animations . . . . . . . . . . . . . .
Drawing for Publication . . . . . . .
Computer Algebra Systems . . . . . .
Non-Mathematical Uses of Geometer

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

30
32
33
34
38
39
40
42

3 Geometric Construction
3.1 The Circumcenter . . . . . . . . . . .
3.2 Available Tools . . . . . . . . . . . .
3.3 The Structure of a Geometer Diagram
3.4 Interpreting Geometer Files . . . . .
3.5 Simple Classical Examples . . . . . .
3.6 Intermediate Examples . . . . . . . .
3.7 Advanced Examples . . . . . . . . .
3.8 Poncelets Theorem Demonstration . .
3.9 Seven Tangent Circles . . . . . . . . .
3.10 A Tricky Construction . . . . . . . .
3.11 Classical Construction . . . . . . . .
3.12 Some Constructions Are Impossible .
3.13 139 More Problems . . . . . . . . . .
3.14 Construction Exercises . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

45
46
47
47
48
49
59
66
67
69
71
73
74
77
77

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

81
82
82
83
84
85
86
89
89
91
92
93
95
96
97
99

4 Computer-Aided Proof
4.1 Finding the Steps in a Proof .
4.2 Testing a Diagram . . . . . .
4.3 Altitudes as Bisectors . . . .
4.4 Another Bisector . . . . . .
4.5 Perpendicular Diagonals . .
4.6 Constant Sum . . . . . . . .
4.7 Find a Radius . . . . . . . .
4.8 Sum of Lengths . . . . . . .
4.9 Bisector Bisector . . . . . .
4.10 The Simson Line . . . . . .
4.11 A Property of Two Cevians .
4.12 Tangent Line Problem . . . .
4.13 Cyclic Quadrilateral Problem
4.14 A Strange Relationship . . .
4.15 Sum of Powers . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

CONTENTS
5 More Useful Theorems

vii
105

5.1

Cevas Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

5.2

Menelaus Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

5.3

Alternate Proofs of Menelaus and Ceva . . . . . . . . . . . . . . . . 110

5.4

Using Menelaus and Cevas Theorem . . . . . . . . . . . . . . . . . 111

5.5

Ptolemys Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

5.6

Herons Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

5.7

The Radical Axis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

6 Locus of Points

121

6.1

Unknown Geometric Locus . . . . . . . . . . . . . . . . . . . . . . . 121

6.2

Eulers Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

6.3

Triangle from Three Medians . . . . . . . . . . . . . . . . . . . . . . 124

6.4

The Ellipse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

6.5

Locus Construction Exercises . . . . . . . . . . . . . . . . . . . . . . 127

6.6

Constant Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

6.7

Higher-Order Curves . . . . . . . . . . . . . . . . . . . . . . . . . . 129

6.8

Plotting Curves of the Form: y = f (x) . . . . . . . . . . . . . . . . . 132

6.9

Plotting Curves in Polar Coordinates . . . . . . . . . . . . . . . . . . 133

6.10 Envelopes of Curves: the Epicycloid . . . . . . . . . . . . . . . . . . 134

7 Triangle Centers

137

7.1

General Properties of Triangle Centers . . . . . . . . . . . . . . . . . 138

7.2

The Gergonne Point . . . . . . . . . . . . . . . . . . . . . . . . . . . 138

7.3

The Nine Point Center . . . . . . . . . . . . . . . . . . . . . . . . . 140

7.4

The Fermat Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

7.5

The Nagel Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

7.6

The First Napoleon Point . . . . . . . . . . . . . . . . . . . . . . . . 144

7.7

The Brocard Points . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

7.8

The Mittenpunkt . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

7.9

A New Center . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

7.10 Additional Triangle Centers . . . . . . . . . . . . . . . . . . . . . . . 152

7.11 Trilinear Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . 153
7.12 Triangle Center Finder . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.13 Searching for the True Center of a Triangle . . . . . . . . . . . . . 157

CONTENTS

viii
8 Inversion in a Circle
8.1 Overview of Inversion . . . . . . . . . . . .
8.2 Formal Definition of Inversion . . . . . . . .
8.3 Simple Properties of Inversion . . . . . . . .
8.4 Preservation of Angles Under Inversion . . .
8.5 Summary of Inversion in a Circle . . . . . . .
8.6 Circle through a Point Tangent to Two Circles
8.7 Circle Tangent to Three Circles . . . . . . . .
8.8 Ptolemys Theorem Revisited . . . . . . . . .
8.9 Fermats Problem . . . . . . . . . . . . . . .
8.10 Inversion to Concentric Circles . . . . . . . .
8.11 The Steiner Porism . . . . . . . . . . . . . .
8.12 Four Circle Problem . . . . . . . . . . . . .
8.13 The Arbelos of Pappus . . . . . . . . . . . .
8.14 Another Arbelos Result . . . . . . . . . . . .
8.15 The Problems of Apollonius . . . . . . . . .
8.16 Peaucelliers Linkage . . . . . . . . . . . . .
8.17 Feuerbachs Theorem . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

161
162
164
165
170
172
173
174
175
176
178
179
180
182
184
185
188
189

9 Projective Geometry
9.1 Desargues Theorem . . . . . . . . . . .
9.2 Projective Geometry . . . . . . . . . . .
9.3 Monges Theorem . . . . . . . . . . . . .
9.4 The Theorem of Pappus . . . . . . . . . .
9.5 Another View of Projective Geometry . .
9.6 Projective Duality . . . . . . . . . . . . .
9.7 Pascals Theorem . . . . . . . . . . . . .
9.8 Homogeneous (Projective) Coordinates .
9.9 Higher Dimensional Projective Geometry
9.10 The Equation of a Conic . . . . . . . . .
9.11 Finite Projective Planes . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

193
193
195
197
199
201
203
204
210
212
223
224

10 Harmonic Point Sets

10.1 What is an Harmonic Set? . . . . . .
10.2 Some Constructions of Harmonic Sets
10.3 Harmonic Points and Music . . . . . .
10.4 The Complete Quadrilateral . . . . .
10.5 Poles and Polars . . . . . . . . . . . .
10.6 The Problem of Apollonius (Again) .
10.7 Poles and Polars Relative to Conics .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

227
227
230
232
233
236
240
241

.
.
.
.
.
.
.

.
.
.
.
.
.
.

CONTENTS
11 Geometric Presentations

ix
243

11.1 General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . 244

11.2 Improving the Appearance of a Diagram . . . . . . . . . . . . . . . . 245
11.3 Building Geometer Proofs and Constructions . . . . . . . . . . . . . 247
11.4 Some Final Touches . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
11.5 Calculation in Geometer . . . . . . . . . . . . . . . . . . . . . . . . 254
11.6 Using Geometric Transformations . . . . . . . . . . . . . . . . . . . 259
11.7 Constructing Geometer Scripts . . . . . . . . . . . . . . . . . . . . . 261
11.8 Scripts and Layer Conditionals . . . . . . . . . . . . . . . . . . . . . 270
11.9 Computer-Generated Diagrams . . . . . . . . . . . . . . . . . . . . . 271
12 Geometer Proofs

279

12.1 Illustrated Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . 280

12.2 Proof Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
A Mathematics Review

293

A.1 Types of Geometric Problems . . . . . . . . . . . . . . . . . . . . . . 294

A.2 Congruence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
A.3 Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
A.4 Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
A.5 Angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
A.6 Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
A.7 Quadrilaterals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
A.8 General Polygons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
A.9 Circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
A.10 Trigonometric Definitions . . . . . . . . . . . . . . . . . . . . . . . . 327
A.11 Coordinate Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . 336
A.12 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
A.13 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
B Geometer Art

355

CONTENTS

x
C Geometric Problem Solving Strategies

359

C.1 Congruence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361

C.2 Similarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
C.3 Special Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362
C.4 Concurrence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
C.5 Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
C.6 Equality and Inequality . . . . . . . . . . . . . . . . . . . . . . . . . 366
C.7 Ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
C.8 Inversion in a Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
C.9 Algebraic Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . 368
C.10 When to Draw the Line . . . . . . . . . . . . . . . . . . . . . . . . . 369
C.11 Construction Techniques . . . . . . . . . . . . . . . . . . . . . . . . 370
C.12 Relabeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
C.13 General Problem Solving Approaches . . . . . . . . . . . . . . . . . 371

Chapter

Introduction
This book describes techniques and strategies that use computer software to enhance
the study of geometry. If you are itching to try the Geometer program on the enclosed
CD, that is as good a way to start as any. Install the program and go directly to the
tutorial which you will find under the Help pulldown menu.
The Geometer program upon which this book is based is freely available and in
the public domain. Anyone may obtain a copy at:
http://www.geometer.org/geometer

You may already be proficient in the use of one of the commercial computer geometry programs such as Geometers Sketchpad, Cabri Geometry, Cinderella or
others, and if so, almost all the techniques described here can be applied using those
packages.
The main advantage of Geometer is that the included CD contains
computer-readable versions of all the examples and figures in this book in Geometer
format.
1

CHAPTER 1. INTRODUCTION

This book is for anyone who wants to learn about computer methods to enhance the
study of geometry, be they students or teachers of the subject. The term student refers
to anyone who is studying geometryeither as part of a formal course or simply for
the love of the subject. The text does assume a high school mathematics background,
but Appendix A reviews all the important prerequisites.
The purpose of this book is to demonstrate techniques, not theorems. Although
as far as possible interesting theorems have been chosen for the demonstrations, no
attempt has been made to assure that the list of theorems presented is complete in any
sense. While it is true that the text material combined with the additional examples
on the CD does cover a large number of theorems, please do not feel slighted if your
favorite theorem does not appear.

1.1 Computer Geometry Programs

A computer geometry program is a computer program with a graphical user interface
that allows you to draw geometric objects and to adjust them dynamically. Such systems are constraint-based in that most of the objects in a figure are not placed absolutely, but are defined in terms of the positions of other objects on the screen. For
example, your drawing may contain two freely movable points and a third point that is
their midpoint. In most computer geometry programs you can freely adjust the positions of the first two points, but not the third. If you do adjust either of the first points,
however, the position of third point will change so that it remains at their midpoint.
The constraints in such a system can be nested arbitrarily deeply. For example, a
line passing through the midpoint described in the previous paragraph will also move
when the midpoint moves, but the midpoint can only move if one of the two original
points is moved, and so on.
This book demonstrates many things that computer geometry programs can do, but
it is equally important to be aware of what they cannot do.
Such programs do provide accurate drawings. They allow you to visualize not just
one example of a drawing, but thousands of variations. They allow you to test geometric conjectures, and can present dynamic illustrations of relationships or theorems.
What they cannot do (at least as of the publication date) is to generate proofs of
theorems. In order for a theorem to be accepted as true, a formal proof is required, and
although computer geometry programs can help you to find that proof, it is you who
will have to provide the completed version.
The more geometry you know, the more helpful a computer can be. If you have
drawn a diagram and see in it a relationship that you would like to prove, it may be that
the computer can discover other relationships in the diagram that you did not notice
and that may help you construct a proof. But it will be up to you to discover how to
use that information to produce a formal proof.
For example, imagine that you are trying to show that three circles in a diagram
meet at a point. When you allow the computer to examine the diagram, suppose it
notices that a certain set of four seemingly unrelated points are concyclic (concyclic

1.2. THE CONTENTS OF THE CD

means that they all lie on the same circle). This unlikely occurrence is easy for a
computer to recognize, but usually difficult for a human, and it could easily be related
to your three circle concurrence problem. But even if knowing that the points are
concyclic makes your result obvious, you still need to show that the four points are, in
fact, concyclic. It is clear that the more ways you know to prove that four points lie on
the same circle, the easier your task will be. (If youre interested in what such a list
might look like, see item 15 in Section 1.3.)
Another obvious limitation of computers is due to round-off error in their calculations. It should not surprise you if a computer indicates that a certain right triangle
has sides of lengths 3, 4, and 5.0000001, although any student of high school geometry knows that this cannot be true. Thus some common sense is required. If you are
asked to find an unknown length, and the computer measures it as 7.000013, you will
probably have a lot more luck if you try to prove that it is 7, not 7.000013. If, however,
the computer says the length is 7.071907, it is probably
not equal to 7. If you are very

good, you may notice how close 7.071907 is to 50knowledge is power.

1.2 The Contents of the CD

The enclosed CD contains a copy of the public-domain computer geometry program
called Geometer that runs on Windows and Linux machines as well as on Macintosh
systems running OS X. All the illustrations in this text were prepared with Geometer.
See Section 1.5 for details about how to interpret the captions on the figures. The
Geometer CD includes:
The Geometer program itself.
A tutorial and users guide for the program.
The source files for all the examples in the users guide and tutorial.
The source files for all the illustrations in this book.
Solutions to all the exercises in this book in the form of Geometer files (called
Geometer diagrams).
Many additional files of interesting geometric results not covered in this book.
The Geometer program can be used at many levels, depending on how much you
want to learn about it. See Chapter 2 for a much more detailed description of what
Geometer (and programs like it) can do.
To view a diagram in this text simply double-click the corresponding Geometer
file on the CD.
To manipulate the figure, place the cursor over a point and press down the mouse
button. Then until you release the button, the point will be dragged with the
cursor, and the figure will be modified appropriately.

CHAPTER 1. INTRODUCTION

Some Geometer diagrams are proofs. The author of the proof organized the
diagram so that you can step forward and back through the proof using the Start,
Next and Prev (previous) buttons. At each stage during the proof, the figure can
be manipulated as described above.
Still other Geometer diagrams are scripts, in which case the Script button will be
solidly drawn. If you click on this button with the mouse, the diagram will pass
through a prearranged script. Sometimes you can manipulate the figure before
pressing the Script button to observe the consequences of the script with different
initial configurations.
Using the mouse buttons and menus you can construct and save your own simple
Geometer diagrams to test conjectures and to search for patterns and geometric
relationships.
In addition to manipulation with the mouse using a graphical user interface, Geometer diagrams have a textual form that can be edited either within Geometer
or using your favorite text editor. Using this technique, all the features are available, including designing your own scripts or proofs.

1.3 Organization of the Book

Here is the basic organization of the rest of the book:
1. Introduction
What you are reading now.
2. Computer-Assisted Geometry
This chapter consists of a quick survey (with examples) of almost every possible
use for a computer geometry program.
3. Geometric Constructions
To be able to use a computer geometry program, you have to know how to
construct diagrams that correspond to the problems you are investigating. This
chapter describes with many examples methods to construct computer diagrams.
Classic straightedge and compass constructions are described and discussed, but
most of the emphasis is on how to use the more powerful techniques available in
computer geometry programs, Geometer in particular.
At the end of the chapter is a list of construction exercises whose solutions appear
in Geometer files on the CD.
4. Computer-Aided Proof
This chapter describes how to use your computer to help you search for a proof
of a geometric theorem.

1.3. ORGANIZATION OF THE BOOK

5. More Useful Theorems

Here are a few useful theorems and their proofs that are not typically covered
in a high school course. When possible, Geometer will be used as an aid to
find those proofs. The results in this chapter plus those in the review appendix
provide all the classical tools needed to solve the rest of the problems in the book.
6. Locus of Points
Computer geometry programs make the search for locus of a point problems
almost trivial. In this chapter we will examine some nice examples.
7. Triangle Centers
The centroid, circumcenter, incenter and orthocenter are four well-known classical triangle centers, but there are hundreds of others. This chapter examines
some of them.
8. Inversion in a Circle
The technique of inversion in a circle is described, with and without computer
geometry programs. Applications of this technique to geometric construction are
addressed.
9. Projective Geometry
Geometer and probably most other computer geometry programs do many of
their calculations using projective geometry rather than Euclidean geometry. In
addition, much of computer graphics is based upon projective geometry. This
chapter presents the fundamental concepts of projective geometry, but from a Euclidean point of view. Calculations in homogeneous coordinates are described.
10. Harmonic Point Sets
This is a continuation of the previous chapter on projective geometry, but again,
from a Euclidean point of view. In many geometric configurations, sets of harmonic points appear, and if that is the case, many methods can be applied to help
solve problems. Geometer is very good at finding harmonic point sets.
11. Geometric Presentations
This chapter is primarily for teachers of geometry or for anyone who would like
to present geometric results to others using Geometer. Various techniques are
discussed that will improve the quality of Geometer diagrams making them easier to understand and more beautiful. Techniques for the presentation of proofs,
constructions, and animated scripts are discussed with examples.
12. Geometer Proofs
Even if Geometer is not used to find a proof, it can be used to present a proof
step by step in a very intuitive way. It requires some effort to construct such
proofs, but if they are already constructed, such proofs are easy to use, both by
a student or a teacher. At the beginning of the chapter is a list of problems with
diagrams; at the end is a list of problems described with text only. Complete
solutions for all the problems appear as Geometer diagrams on the CD.

CHAPTER 1. INTRODUCTION

6
13. Appendix: Mathematics Review

Here we review parts of high school mathematics. Most of the topics are from
geometry, but some trigonometry and analytic geometry are included. The important results are listed, where important means that the result is useful for
deriving other results. The results are standard, and most can be found in high
school mathematics texts.
14. Appendix: Geometer Art At the beginning of each chapter is a Geometer illustration that is at least partly artistic. This appendix describes how each one is
made, and perhaps says something about the mathematics illustrated.
15. Appendix: Geometric Problem Solving Strategies
This appendix categorizes many of the standard methods to show properties of
geometric figures.

A
B
C

Figure 1.1: When do four points lie on a circle?

Intro/FourCircPts.D [D]

For example, to continue with the example in Section 1.1, how does one show
that four points lie on a circle? Here are some approaches if the points are,
in clockwise or counterclockwise order, A, B, C, and D (see figure 1.1). (Do
not worry if these techniques seem mysterious. They will be covered in Appendix A.)
Show that all four points are the same distance from some fixed point
(which would be the center of the circle).
Show that DAB = BCD = 90 (or that ABC = CDA = 90 ).

More generally, show that DAB + BCD = 180 (or that ABC +
CDA = 180 ).
Show that ABD = ACD (or that BCA = BDA or CDB =
CAB or DAC = DBC).
Show that all four points lie on some well-known circle, such as the nine
point circle.
(Ptolemys Theorem) Show that AB CD + BC DA = AC BD.

The entire appendix is organized a bit like a thesaurus. You look up the sort
of problem you are attempting to solve and you will find a list of possible approaches to that kind of problem.

1.4. PROOFS IN THIS BOOK

1.4 Proofs in this Book

We have a habit in writing articles published in scientific journals to
make the work as finished as possible, to cover up all the tracks, to not
worry about the blind alleys or describe how you had the wrong idea first,
and so on. So there is not any place to publish, in a dignified manner, what
you actually did in order to get to do the work.
Richard P. Feynman
I have had my results for a long time: but I do not yet know how I am
to arrive at them.
Karl Friedrich Gauss
Most geometry texts present slick and polished proofs of each theoremusually
the best one that the author knows. In virtually every case, the first discovered proof
of a new theorem is not slick and polished. The slick and polished proofs arise as
mathematician after mathematician improves upon what has been done before.
There are different kinds of proofs. Some are seemingly miraculous, and it is almost
impossible to imagine how they were discovered. Some are beautiful in a different way
in that they make it completely obvious why a certain fact or theorem is true. Some
are mechanical, and relatively easy to construct, although they may not be brilliant and
may not shed too much light on the problem.
One of this books goals is to show how to use computers to help solve problems.
In geometry this often means you are trying to demonstrate some property that you
have never proved before. Thus the proofs here will tend to be of a rougher form,
more mechanical and less magical than what you will find in geometry textbooks. The
advantage to this is that it is often much easier to see how to arrive at that proof.
Usually we will not concern ourselves with the process of polishing the proof to
make something beautiful, but from time to time there is such a wonderful and magical
proof available that it would be aesthetically criminal to omit it.

1.5 Illustrations in this Book

Uninvited advice is usually ignored; yet I wish to offer some in the
hope that it will be helpful. No book on mathematics can have enough
illustrations or formulas. The thorough reader must always work with
pencil and paper at hand.
Nicholas D. Kazarinoff
Although Kazarinoffs claim from [Kazarinoff, 1961] is true, illustrations take a
lot of space so many books have fewer than they should. The author remembers how
depressed he was once when he opened a book whose title claimed that it presented the
geometric viewpoint of complex variables, and there were almost no illustrations.

CHAPTER 1. INTRODUCTION

But with a computer that has decent graphics and a suitable geometry program,
it is possible to present not just one illustration of a geometric fact, but thousands or
even millions of variations of that illustration. The final sentence of Kazarinoffs quote
should be amended to work with pencil, paper, and computer at hand. (By the way,
his book ([Kazarinoff, 1961]) is pretty good too.)
This book contains quite a few illustrations, and every one of them is also available
as a Geometer file that can be viewed and modified dynamically. In addition to being
simply a displayer of these geometric diagrams, however, Geometer allows both the
modification of existing diagrams and the creation of new ones.
Following each figures caption is a line that tells you where to find the Geometer
diagram used to produce the figure, including the directory and file names. This name
is relative to where you installed Geometer. All Geometer file names have a suffix
of .T or .D. There is no difference between the two file types from Geometers
point of view, but if the figure cannot be reasonably manipulated and is only used as a
drawing, then the .D suffix is used. Think of the .D as standing for drawing and of
.T as standing for theorem.
Following the file name in brackets is a single letter that indicates what sort of a
diagram it is. There are four possibilities:
D Drawing. These are the least interesting and usually have the .D suffix. A simple drawing was needed for the text and although some parts may be movable, if
you move them, the figure may no longer illustrate what it was supposed to. But
even these figures can be interesting. If you are wondering how on earth some of
the diagrams were drawn with Geometer, you can simply load the diagram and
use the Edit Geometry command to see. (Or examine the .D file in your favorite
text editor.)
M Manipulation. The drawing can be manipulated by moving various points.
Usually it is obvious from the text what parts should be moved, but if not, the
diagram itself will usually contain some clues in the form of text labels. In a
tiny percentage of these manipulation diagrams there are some extra labels that
are not tied to the geometry correctly but are placed where they are to make a
diagram suitable for publication. The drawing changes correctly, but sometimes
the labels are left behind.
P Proof. When you see this sort of diagram, it is almost always a good idea to
load it into Geometer. It is either a proof or construction where you are led,
step by step, to the solution. To advance to the next stage, use the Next button
in the control area of the screen, or simply type the letter n on your keyboard.
To go back, use the Prev button or type p on your keyboard. To restart the
proof or construction, use the Start button. In almost every proof, you can also
manipulate the figure, both before or during the proof.
S Script. A script is a Geometer file that includes a scripted transformation of
the drawing. These diagrams are also worth loading into Geometer. The builtin script can be accessed by pressing the Run Script button. To interrupt the

1.6. HOW TO USE THIS BOOK

script, simply click the mouse button with the cursor anywhere in the Geometer
window. A few of the scripts do not do too muchthey were simply used to
guarantee evenly-spaced points for diagrams for publication, but the majority
are very useful to watch. In most cases, you can also manipulate the figure
before running the script. For example, a script that draws a conic section given
information about certain lengths and foci can be run, then the lengths and/or
foci changed, and when the script is run again you can see how the modifications
affected the result.
If a picture is worth a thousand words, then are a thousand pictures worth a million
words? Even if not, a thousand pictures is surely better than one. In addition, the fact
that on a computer screen you can watch each of the thousand images blend into the
next tends to provide an even better idea of what is going on.
Another less obvious advantage of using Geometer is that since every illustration
in this book is also a Geometer file on the CD, you can view and manipulate that
image on your computer screen as you read about it and it is easy to continue viewing
the diagram when you turn the page.
In a few cases, there are two or more versions of the Geometer file on the CD,
since a slightly different version was used to generate the figure in the text. A common
situation is when the basic figure appears on the left and another version with some
additional constructions appears on the right. The file whose name appears with the
illustration is usually neitherit is usually the basic figure plus the constructed lines.
Finally, on the first page of each chapter there appears an illustration that is more
artistic than the usual figure in a Euclidean geometry text. Appendix B provides a short
description of each, and the name of the Geometer file and (if appropriate) the name
of the C file that generated that Geometer file.

1.6 How to Use this Book

One of the reasons Americans on average do not do well in mathematics is that nowadays we are conditioned by television and our society in general to expect instant gratification. Unfortunately, whether you like it or not, the best way to learn anything about
mathematics is to think hard about it and to struggle with problems. If you simply read
a problem and then immediately read its solution, you will not get too much out of it.
The more you struggle with a problem, even if you do not solve it, the more you will
learn. If you do solve it by yourself, great! If you needed a hint or two, that is great
too. But even if you get totally stumped, at least you will read the solution with much
more interest.
Much of this book is organized as as follows: A problem is presented followed by
a section that discusses how Geometer might be used to search for approaches to it.
Finally, a complete solution to the problem is presented. You will learn the most, of
course, if you read the problem and try to use Geometer yourself to try to solve it. See
what you can learn. Only then should you look ahead at the solution.

CHAPTER 1. INTRODUCTION

10

Chapter 3 ends with a long set of construction problems. Each has a complete
solution in a Geometer file on the CD, but of course there is no discussion of how that
construction was imagined. Similarly, Chapter 12 consists entirely of problems, each
of which has a corresponding proof in a Geometer file.
As you work on each problem, try not to get stuck in a rut. Draw pictures (even
better, draw pictures with Geometer). Load the associated Geometer file, if there is
one. Look at the solution in the Geometer file only after you have worked on the
problem yourself, whether you solved it or not. Even if you did solve the problem,
check the Geometer solutionit may be different from yours, and may shed additional
light on the problem.

1.7 Notation
A foolish consistency is the hobgoblin of little minds, adored by little
statesmen and philosophers and divines. With consistency a great soul has
simply nothing to do.
Ralph Waldo Emerson
 For fools rush in where angels fear to tread.
Alexander Pope
 For mathematicians rush in where fools fear to tread.
Tom Davis
Some topics in this text are more difficult than others. Even within a topic, some
paragraphs are difficult and can be skipped on the first reading. Others are very difficult,
and only experts (or angels or fools) should wander into them on the first pass.
Difficult and very difficult paragraphs are indicated with black diamonds that look
like this:  and this: . If you are a skier, you will recall that the most difficult slopes
are often marked double black diamond. Sometimes entire sections are marked with
a single or double black diamond in which case everything in the section has the indicated difficulty. You may even encounter a triple black diamond: . If an entire
section is marked with a warning and if some subsection or paragraph within it is also
marked with a warning, it means that the subsection is relatively more difficult than the
section within which it lies.
This books mathematical notation is fairly standard, but following Emerson, it is
not 100% consistent.
We use the symbol
= to indicate congruence between two equal geometric figures. The symbol = is for equality of lengths, angle measure, area, et cetera. The
symbol indicates two similar figures.
Generally, points or vertices are indicated with upper-case letters, as: A, B, . . . Z.
When they have a name, lines or line segments are indicated with lower-case letters:

1.8. WHERE TO GO FROM HERE

11

a, b, . . . , z. A segment or line can be indicated by two points on the line, as: The line
AB.
4ABC is the triangle whose vertices are the points A, B, and C. For more complex polygons, we will normally just list the vertices and include a word to say what
kind of polygon we mean, such as: quadrilateral ABCD or hexagon ABCDEF .
The same notation is used for a segment as for the length of the segment, and that
should
not cause any confusion. If it appears in some equation, such as AB : CD =

2 or AB + BC AC, we are obviously talking about the lengths of the segments.

Ratios of numbers, lengths or areas are often written as follows: a : b = c : d. This

is almost equivalent to a/b = c/d, but remains true if b = d = 0, as long as a and c are
not zero.

If angles have a name, it will usually be a Greek letter, like , , , . . .. If there

is no name, the notation: ABC means the angle traced out by drawing a line from
point A to point B and then on to point C. Sometimes the name of the vertex may
be used as the angle name if there is no possibility for confusion. For example, in
4ABC, if there are no other lines passing through vertex A, then A is the same as
BAC.

If two lines or segments AB and CD are parallel we write AB k CD. If those

same two segments were perpendicular, we would write AB CD.


If two points A and B lie on a circle, the symbol AB refers to the arc from A to B.
If there is any chance of confusion, the arc is the part of the circle beginning at A and
going to B in a counterclockwise direction, so the arcs AB and BA together make up
the entire circle.
We will use the symbol A to indicate the area of a polygon. A(4ABC) stands for
the area of 4ABC.

A few sets are important. N = {0, 1, 2, 3, . . .} represents the set of natural numbers.
Some people include zero in the set of natural numbers and some do not. In this text,
we will include it. Z = {. . . , 4, 3, 2, 1, 0, 1, 2, 3, . . .} is the set of integers. Q
is the set of rational numbers, R is the set of real numbers, and C is the set of complex
numbers.

1.8 Where to Go from Here

The main reason to read this text is to learn some computer techniques that are useful
in Euclidean geometry, but in the process of learning those techniques you will be
exposed to quite a few interesting and beautiful theorems that do not appear in high
school textbooks. If you want to learn more geometry or have more problems to try,
there are plenty of other sources available.
One computer technique not mentioned up to now is simply the availability of a
great deal of information on the internet. Here is a short list of some sites that deal with
various aspects of geometry. Unlike standard bibliographic references, it is impossible
to guarantee that the addresses below are still valid when you read this book.

12

CHAPTER 1. INTRODUCTION
http://www.geometer.org/geometer/
The Geometer home page. Here you can obtain the latest version of the Geometer program. There is additional material here of various sorts, including
additional Geometer diagrams.
http://freeabel.geom.umn.edu/docs/forum/
The Geometry Forum. This is an electronic community focused on geometry
and math education based at Swarthmore college.
http://www.ics.uci.edu/~ eppstein/junkyard/
The Geometry Junkyard. A collection of usenet clippings, web pointers, lecture notes, research excerpts, papers, abstracts, programs, problems, and other
stuff related to discrete and computational geometry.
http://www.geom.umn.edu/
The Geometry Center. Although the Geometry Center at the University of
Minnesota is now closed, its website is still available and filled with interesting
items.
http://dmoz.org/Science/Math/Geometry/
Open Directory Project Science: Math: Geometry. An index to many more
interesting web pages related to geometry.
http://www.cut-the-knot.org/ctk/index.shtml
Cut The Knot! An internet column by Alex Bogomolny with many wonderful
articles mostly about geometry, and many with associated animations.
http://faculty.evansville.edu/ck6/tcenters
Encyclopedia of Triangle Centers. Clark Kimberlings collection of over 1000
triangle centers. See Chapter 7.

Finally, here is a list of some books that may be interesting. See the bibliography
for complete citations.
Geometry Revisited ([Coxeter and Greitzer, 1967]) is one of the best books on
general advanced Euclidean geometry. The information is quite dense, and it is
amazing how much is packed into one small book.
College Geometry ([Altshiller-Court, 1952]) is an old classic textbook that is also
loaded with information. It contains an extensive set of problems and can often
be found in its paperback version in used bookstores.
Advanced Euclidean Geometry ([Johnson, 1929]) is another classic. It is a bit
more difficult to find than the volume above, but it is worth the search. It was
also more recently (1960) printed in a Dover Publications edition.
A Survey of Geometry ([Eves, 1965]) is yet a third classic worth checking.

1.8. WHERE TO GO FROM HERE

13

Geometry Turned On! ([King and Schattschneider, 1997]) contains a collection

of papers on the use of computer geometry programs both in and out of the
classroom.
Geometric Inequalities ([Kazarinoff, 1961]) is a good place to learn about inequalities. The topic of geometric inequalities is almost completely ignored in
the book you are holding.
Geometric Transformations I, II, and III These three books by Irving Yaglom
([Yaglom, 1962a], [Yaglom, 1962b], [Yaglom, 1962c]) approach geometry from
the point of view of transformations. This is another topic barely touched upon
in the current text.
Dan Pedoe has written a couple of interesting books: Circles: A Mathematical View ([Pedoe, 1995]), as advertised, tells a lot about circles. Another of his
books, Geometry, A Comprehensive Course ([Pedoe, 1988]) describes in great
detail vector methods to solve geometric problems. This second text is a Dover
republication of his A Course of Geometry for Colleges and Universities published in 1970 by Cambridge University Press.
Bergers texts: ([Berger, 1987a], [Berger, 1987b] and [Berger et al., 1984]), including the problem book have their good and bad points. On one hand, they
are wonderful to look throughyou will get hundreds of ideas just by looking at
the illustrations. On the other hand, this author finds the proofs and explanations
very difficult to followBerger seems to think about geometry in a different
way. If you have trouble following this book, perhaps you think like Berger, and
his books will make more sense to you.
Two books called Proofs Without Words ([Nelson, 1993]) and Proofs Without
Words II ([Nelson, 2000]) provide interesting geometric methods to visualize
proofs from many areas of mathematics, geometry included.
Challenging Problems in Geometry ([Posamentier and Salkind, 1988]) contains
a set of problems that are nice and not too difficult. A more challenging set can
be found in the Russian book Problems in Plane Geometry ([Sharygin, 1988]).
Both contain a section of problems, another section of hints, and then a section
of solutions.
Two fun books include Excursions in Geometry ([Ogilvy, 1969]) and The Penguin Dictionary of Curious and Interesting Geometry ([Wells, 1991]). Both contain lots of topics, many of which are a bit off the beaten path.
For bigger view of geometrynot just Euclidean, but the whole gamuttake
a look at Coxeters Introduction to Geometry, Second Edition ([Coxeter, 1989])
and Hilberts Geometry and the Imagination ([Hilbert and Cohn-Vossen, 1983]).
A more rigorous approach to geometry can be found in Euclidean and NonEuclidean Geometries, Third Edition ([Greenberg, 1993]), and in Companion to
Euclid ([Hartshorne, 1997]), or Elementary Geometry from an Advanced Standpoint ([Moise, 1963]).

14

CHAPTER 1. INTRODUCTION
The CRC Concise Encyclopedia of Mathematics ([Weisstein, 1999]) contains an
amazing collection of mathematical information (including much geometry).
Although there is probably no hope of finding these particular books, ancient
high school and college geometry texts (leather-bound, even) can often be found
in used bookstores. They often present a very different view of the subject from
what is usually seen in modern texts. Two examples are New Plane and Solid Geometry ([Beman and Smith, 1900]) and Geometrical Problems Deducible from
the First Six Books of Euclid ([Bland, 1819]), but there are plenty of others. Note
that there is no error in this last citation; Blands book was really published in
1819, not 1919.

Chapter

Computer-Assisted Geometry
There are many computer programs that aid in the visualization of Euclidean geometry. Some available commercial ones at the time of this writing include Geometers
Sketchpad, Cabri Geometry, and Cinderella.
The CD accompanying this book contains yet another called simply Geometer
that provides many of the features of the programs above, but is in the public domain.
Geometer runs on PCs, Macintoshes running OS X and Linux machines. All the examples in this book were created with Geometer, but most of them could have been
drawn equally well with any of the other commercial programs. Similarly, every illustration is a Geometer diagram, so if you read this with a computer at your side so you
can manipulate the Geometer diagram at the same time that you read about it. (The
drawings of geometric figures produced by Geometer are called diagrams.)
To understand this chapter you need to have some idea of how Geometer or some
other computer geometry program works. If you use Geometer, the CD contains complete documentation including a tutorial. If you have never used a computer geometry
15

16

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

program before, it is probably worthwhile to work your way through the Geometer
tutorial before reading too much farther.
Finally, there is a chapter in the Geometer reference manual on the CD that contains another list of suggestions for effectively using the program to study geometry
and to learn to prove theorems.
In each section of this chapter a different strategy or technique that uses a computer
geometry program is discussed, together with one or two examples. Some uses are
obvious, but in certain cases the example is followed by a more detailed technical
description. You will not miss much if you skip the technical details on first reading,
but you may find it valuable to return to them when you try to apply that technique to
your own problem. Remember too that you can always load any Geometer file into a
text editor and discover exactly how it was constructed.

2.1 Accurate Drawings

A computer geometry program can produce quite accurate pictures. If you have tried
to make drawings with a straightedge and compass, you know how hard this isthe
three lines that are supposed to intersect at a point in fact form a little triangle that has
eighth-inch sides, or perhaps the line that is supposed to be tangent to a circle clearly
misses or cuts off a sizeable chunk.
The more complicated the drawing, the more the errors compound themselves.
Most people agree that it is almost impossible to draw, with a physical straightedge and
compass, a reasonable approximation of the construction of the regular heptadecagon
(17-sided figure). On a computer screen, the construction is easily accurate to the
nearest pixel.
This section contains three examples. The first illustrates an interesting relationship
between a pair of constructions. It is simple enough that you can do it by hand with
a pencil and paper and then compare your results with the illustration in the book (or
on your computer screen). The second example shows a situation where a hastilydrawn figure may lead to an incorrect result which could not happen if a computer
geometry program were used. Finally, the third example demonstrates the construction
of a regular heptadecagon which is so difficult that it is virtually impossible to do
without a computer drawing program.

2.1.1 Harmonic Point Sets

For now, do not worry about what is meant by the term harmonic point sets; that will
be covered in Chapter 10.
You may find it interesting, however, to try to duplicate the drawing in figure 2.1
using a straightedge and compass. The difficulties are best illustrated if you do not look
at the figure, but simply read and follow the directions in the following paragraph. But
you will probably find the construction difficult even with the figure at hand.

2.1. ACCURATE DRAWINGS

17

C
E
F
D

Figure 2.1: A Set of Harmonic Points

Computer/Harmonic.T [M]

Construct two unequal circles centered at A and B that neither intersect nor lie one
inside the other. Draw the two lines that are the common internal tangents and label
their intersection I, and similarly, draw the pair of common external tangents and
label their intersection O. Now choose any point C not on the line IO and draw
the segments CO, CA, and CI. Select any point D on the segment CA and draw the
rays ID and OD that intersect the lines CO and CI at points E and F , respectively.
Seemingly miraculously, the line EF passes through the point B, and this will be true
no matter where you choose to place the points C and D, and independent of the sizes
and positions of the original circles.
When you are done, be sure to load the Geometer diagram for figure 2.1 and manipulate the sizes and locations of the circles as well as the positions of the points C
and D.

2.1.2 Obviously True, but False

Here is a classic example found in some high school geometry texts that demonstrates
the dangers of having a proof rely on a drawing. This sort of thing is a big danger for
hand-drawn figures, but it becomes far less of a problem when a computer geometry
program is used.
Although figure 2.2 was drawn with Geometer, it was not drawn using the built-in
constraints, and is, in fact, wrong, as we shall see. It is, however, very much like a
drawing that might be made by hand.

18

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

D

E
Figure 2.2: Bogus Proof
Computer/Incorrect.D [D]

Start with an arbitrary rectangle ABCD as in figure 2.2, and using a compass,
draw an arc of a circle centered at C and passing through B, that goes a little outside
the segment BCto point E.
Since ABCD is a rectangle, ADC = BCD = 90 . Since BCE is a little
bigger than zero, DCE is a little bigger than 90 . We will prove that ADC =
DCEsomething that is obviously not true.
Here is the bogus argument. Construct the perpendicular bisectors of the segments
AB and AE. Since those perpendicular bisectors are not parallel, they will meet at
some point O.
Since ABCD is a rectangle, the perpendicular bisector of AB is also the perpendicular bisector of DC, so all the points on it are equidistant from D and C. Therefore
DO = CO. By similar reasoning, O is equidistant from A and E, so AO = EO.
Points B and E are on the same circle centered at C, so BC = CE, and since
ABCD is a rectangle, AD = BC = CE.
To summarize, DO = CO, AD = CE, and AO = EO. Therefore the two
triangles 4ADO and 4ECO are congruent since they share three equal pairs of sides
(using SSS congruence). Therefore ECO = ADO, and we can subtract the equal
angles CDO and DCO to obtain the result we wantthat ADC = ECD.
The result is clearly not true, but every step seems correct. What is wrong? The
answer is that the diagram is not drawn accurately. (In fact, the author cheated and had
to misuse Geometer to get the desired misleading effect.)
Figure 2.3 shows an accurately drawn diagram of the situation, and it is instantly
obvious what went wrongthe line segment EO lies on the outside of the rectangle.
In fact, if you check in the correct figure, ADO = ECO, which is what we proved.

2.1. ACCURATE DRAWINGS

19

E
Figure 2.3: Correct Figure
Computer/Correct.T [M]

The point is that an accurate figure can correct some very mysterious problems, and
computers allow you to draw extremely accurate figures without too much effort.

2.1.3 Construction of a Regular Heptadecagon

Here is a construction due to Richmond of the regular heptadecagon (17-sided figure).
Geometer produces figure 2.4. See [Conway and Guy, 1996] for more details of this
construction.
To truly drive home the advantages of a computer geometry program over a drawing
with pencil, paper, straightedge and compass, you might try (even while looking at
figure 2.4) to do the construction yourself as described below.
Even with the figure and the text that describes the construction, you will see that
the construction is a bit difficult to follow. If you load the Geometer diagram for this
figure (found in the directory Computer/Heptadecagon.T on the CD accompanying this
book) you can go through the construction one step at a time by clicking repeatedly on
the Next button in Geometer (or pressing the n key). This will make the construction
much easier to follow.
Begin with a circle in which the regular heptadecagon is to be inscribed. Find its
center, and construct horizontal and vertical diameters passing through N , S, E, and
W at the north, south, east, and west edges, respectively.
By finding the midpoint of N O and then the midpoint between that new point and
O again, construct a point A that is 1/4 of the way from O to N .
Construct the circle centered at A passing through E and use that to bisect the
angle EAS twice so that SAB = EAS/4. This new angle will cut the horizontal
diameter EW at B.
Now construct at A a 45 angle relative to AB intersecting EW at C. In other
words, construct C such that BAC = 45 .

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

20

N 5

7
3
8
2

D
A

9
W

O B

1E

10
F

17

11
16
12
13

15
S 14

Figure 2.4: Construction of a regular heptadecagon

Computer/Heptadecagon.T [P]

Find the intersections D nd F of the circle having EC as diameter with the line
N S.
Construct a circle with center B passing through D (or F ), and construct tangents
to it that are parallel to N S (this takes a few stepsfirst find the intersections of the
circle with EW , and then construct the tangents there).

If we call the point E 1, and label the intersections of the vertical tangent lines
from the last paragraph with the original circle 4, 6, 13, and 14 as shown in
the figure, we have 5 points on the heptadecagon, and it is easy to construct the rest.
For example, we could bisect the arc 46 to find point 5, and once we have the arc 45
we can copy it all the way around the circle.

2.2. DRAWING MANIPULATION

21

2.2 Drawing Manipulation

Once a figure is drawn in a computer geometry program, it can be manipulated. If you
draw one figure, it may be that some amazing relationship holds just by chance, but
when you manipulate the diagram, it is extremely unlikely that the chance relationship
will continue to hold. This is particularly true for students in their first geometry course.
When told to draw a triangle, they will almost invariably draw one that is very nearly
equilateral, and for equilateral triangles, all sorts of amazing relationships hold.
In this section we will examine a result that is not as well-known: Miquels theorem.

2.2.1 Miquels Theorem

B
D
A

B
D
E

F
C

B
D
D
A

E
E

A
F

Figure 2.5: Miquels Theorem

Computer/Miquel.T [M]

Given any triangle 4ABC, choose points D, E, and F on the line segments
AB, BC, and CA, respectively. Now, draw three circles through the sets of points
{A, D, F }, {B, E, D}, and {C, F, E}. Notice that all three circles seem to pass
through the same point. (See the triangle in the upper left of figure 2.5.)
Is this just chance? Try moving the points A, B, C, D, E and F to obtain the other
three drawings in figure 2.5. In every case, the three circles seem to pass through a
common point, although that point may not lie in the interior of the triangle 4ABC
(see the example in the lower left), and it even seems to work if the points D, E, and
F are anywhere on the lines: D does not have to be between A and B, for exampleit
just needs to be somewhere on the line AB. (See the example in the lower right.)
This result, called Miquels theorem, is amazing in that it seems to work no matter
what the shape of the triangle, and where the only constraint on the other three points is
that they be constrained to the appropriate lines. Well, not quite anywhere: to determine
a proper circle, the points A, D, and F must be distinct, et cetera. But the theorem can

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

22

even be extended to make sense in that case. If A and F are the same, for example,
then the circle passing through A, D, and F can be replaced by the circle tangent to
the line AC at A and passing through D.
Load the Geometer diagram from Computer/Miquel.T and manipulate it to check
other configurations, to experiment with the degenerate cases, and generally to amaze
yourself with the beauty of Miquels theorem.

2.3 Finding and Testing Conjectures

2.4 conjecture testing
Accurate drawings allow you not only to test conjectures, but to discover new ones.
The various different configurations of Miquels theorem in Section 2.2.1 above show
us that not only does the theorem seem to hold since the circles always seem to intersect
in a point, but that the theorem is perhaps more general than we first imagined. It seems
to hold, for example, even if the points on the sides of the triangle are outside what
we normally consider to be a triangle.

2.4.1 Morleys Theorem

Here is another wonderful example of the same sort of thing, based on Morleys theorem.
C

Z
A
C

B
X

Figure 2.6: Morleys Theorem

Computer/Morley.T [M]

Morleys theorem states that if all three of the angles of an arbitrary triangle are trisected as in the diagram on the left in figure 2.6, then the intersection points of adjacent

2.4. CONJECTURE TESTING

23

pairs of trisectors form an equilateral triangle. In the figure, 4XY Z is equilateral, and
4ABC is arbitrary.
This is extremely interesting, and when you adjust the Geometer diagram, it seems
to continue to work. But even more interesting is what happens if you turn the triangle
4ABC inside out, as in the diagram on the right of the same figure. (In other words,
make the vertices A, B and C go clockwise instead of counterclockwise.) The labeling
is the same, and the triangle 4XY Z continues to be equilateral, but at first glance, the
angles of 4ABC no longer seem to be trisected. On closer inspection, however, they
are trisected, in the sense that it is now the exterior angles that are trisected. (Remember
that the trisectors are lines and the figure only shows a segment of the trisecting lines.
Extend them in the opposite direction and it will be more obvious that they are the
trisectors.) Convert the trisector lines from segments to infinite lines and it is clearer,
perhaps, that they remain angle trisectors.
Thus Morleys theorem holds if all the trisectors are of interior angles or all are
trisectors of exterior angles. What if two are interior and one exterior? What if two are
exterior and one interior? Try ityou will probably be surprised at what you learn.

2.4.2 New Triangle Centers

Quite often, if you draw a figure with some
C
B
sort of symmetry in the construction, you
A
will find that interesting relationships hold.
C
B
Thus if you are trying to make up problems
for a math class or examination or contest,
symmetric doodling will often generate
B
C
interesting relationships.
A
As a quick example, let us find a couple of triangle centers. The most familiar
A
centers are the centroid (where the three
medians of a triangle intersect), the incenter (the intersection of the three angle bi- Figure 2.7: An Unknown Triangle Center
sectors), the circumcenter (the intersection
Computer/Tcenter.T [D]
of the three perpendicular bisectors of the
sides) and the orthocenter (the intersection of the three altitudes). All are interesting
because in general, three lines do not meet at a point.
It is surprisingly easy to find other triangle centers, and in fact, all of Chapter 7 is
devoted to doing so. But figure 2.7 illustrates a simple one.
In 4ABC, each vertex is reflected across the opposite side resulting in vertices
A0 , B 0 and C 0 . If the centers of the circumcircles of 4A0 BC, 4AB 0 C and 4ABC 0
are connected to the opposite vertex as in figure 2.7, all three of those lines intersect at
a point which will be a new triangle center. (Notice also that the three circumcircles
also seem to intersect at a point. Proving either of these concurrences might be suitable
problems for a math exam or contest. Of course you had better make sure you know
how to solve them before you assign them as problems. They might be very difficult to
prove.)

24

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

2.5 Ease of Correction

It is easy to erase your mistakes on a computer screen. This is probably most useful
when you accidentally draw the wrong thing, but there are other uses as well.
For example, if you suspect that four
points in a diagram lie on a circle, just
draw a circle through three of them and
see if it goes through the fourth. If it
does, manipulate the figure a bit to see if
the circle went through that fourth point
by chance. In either case, you can just
erase the circle afterwards if you want
to. If you are working with a pencil and
paper, you will probably have to draw
a couple of extra construction lines just
to draw the circle, so erasing it is more
difficult. Besides, if it does go through
Figure 2.8: Monges Theorem
the fourth point and you want to see if
Computer/Mongedemo.T [D]
that occurred by chance, you may have
to redraw the entire diagram with different measurements.
Here is a very simple example where the ability to erase might be useful. Draw
three unequal circles where none of them intersect and none of them lies completely
within another. For each of the three pairs of those circles, draw the common external
tangents and identify where those tangents meet. It appears that those three points lie
on a line. Is it true? Draw a line through two of them and see if it passes through the
third. If so, manipulate the figure and see if it continues to hold. See figure 2.8.
With a physical paper and pencil, lots of auxiliary lines must be drawn to construct
the common tangents, and if you wish to test the theorem in different configurations,
then far more erasure would be required.
The intersections do indeed lie on a line. The result is known as Monges theorem
and it is proved in Section 9.3.

2.6 Stepping through Proofs

Both teachers of geometry and students may find this feature valuable. The proof of
many interesting theorems requires a number of steps and construction lines, and if a
student (especially a beginner) is faced with a complex diagram covered with additional
construction lines, it is difficult to look through the diagram and see why 4ABC
=
4DEF . The student must find the points A, B, . . . , F in a possibly very complex
diagram and then notice that the two triangles determined by them are congruent.
If a Geometer diagram is constructed as a Proof, then when the student steps
through the proof and reaches the statement that 4ABC
= 4DEF , both 4ABC
and 4DEF will be highlighted. In addition, the student may then manipulate the

2.6. STEPPING THROUGH PROOFS

25

diagram at this point and see that the two triangles remain congruent under all the
manipulations. When the student continues to the next step of the proof, the subject of
that step will be highlighted, and the highlighting for the previous step will disappear.
Teachers of geometry can prepare, with a little bit of effort, proofs of theorems that
can be displayed step by step on a computer screen. The actual construction of such
proofs is beyond the scope of this book. To learn the techniques for constructing your
own proofs using Geometer, see the reference manual, especially the chapter entitled
Teachers Tutorial, where about 35 pages of text are devoted to this topic.
Fortunately, the CD contains many Geometer diagrams in the form of proofs, for
theorems ranging from those covered in a typical high school text to the more interesting advanced theorems covered here. Most of them can be used in a classroom without
modification, so teachers of geometry can use them without learning to construct them.
The many proofs on the CD also provide good examples for a teacher who wishes to
construct her own.
However, it is worth reading through the example below, since it does illustrate the
sorts of things that can be done to make it easier to follow a proof on a computer than
on a printed page of a book.
We will use the nine-point circle theorem to demonstrate Geometers ability to
present proofs. If you can use a computer, read the rest of this section as you view the
diagram on your screen. Load the file Computer/Ninepoint.T and you should see an
image on your screen roughly like figure 2.9, but with some additional accompanying
text.
The statement of the theorem is this:
Theorem 2.1 (The Nine-Point Circle) Given any triangle, the three midpoints of the
sides, the three feet of the altitudes, and the three points midway between the orthocenter and the vertices all lie on the same circle, called the nine-point circle.
In figure 2.9, let 4ABC be the triangle, the midpoints of the sides are A 0 , B 0 ,
and C 0 , the feet of the altitudes are E, F , and G, the orthocenter (the point where the
three altitudes meet) is H, and the points Q, R, and S are midway between H and the
vertices A, B, and C. All nine points: A0 , B 0 , C 0 , E, F , G, Q, R, and S lie on a circle.
With a printed version of the diagram, you are stuck. You can see that it works
there, but that is it. On the computer screen, you can move points A, B, and C, and
watch how all nine points move in response and how, although it changes size and
position, there is always a single circle that passes through the points. You see that it
remains true whether the feet of the altitudes lie inside or outside the triangle. You see
how for some triangles, the circle is enormously larger. You see that the ordering of the
points may change on the circle, but still the theorem holds.
There is more, of course. The illustration in the text is in black and white. On the
computer screen, the altitudes and their feet are in red, the midpoints of the sides are in
yellow, and the other points Q, R, and S, are in green. Finally, the text on the screen
restates the theorem, suggests things you might do to test it, and finally, it says, Press
Next to continue . . . .

26

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

Each time you click on the Next button (or press the n key on your keyboard), the proof advances by one step.
B
The text under the illustration explains
and justifies the step, and the object
R E
F
or objects described by the text can be
H
highlighted in some way. For examC
A
ple, the first step of this nine-point cirQ
S
cle proof is to show that 4ABC
4C 0 BA0 . The text explains why this is
A
G B
C
the case, but in addition, the outlines of
those two triangles blink in different colFigure 2.9: The Nine-Point Circle
ors. This is not so critical in a diagram
Computer/Ninepoint.T [P]
as simple as this, but it is easy to imagine a complex figure where there are so
many points and lines that it is difficult even to find the triangles mentioned. When you
advance to the next step of the proof, the colors of the triangles return to normal, and
the features important in the next step are somehow highlighted, et cetera.
Construction lines can appear and disappear at different steps of the proof so as to
avoid cluttering the diagram. They appear only when needed.
Another advantage of a Geometer presentation is that at any stage in the proof,
you can manipulate the points of the diagram. In this proof a nice example occurs in
the third step where the segments A0 C 0 and SQ are shown to be parallel. As you look
at that step, you can move the point B up and down and notice not only that the lines
remain parallel, but that they can swap positions. So if your proof, for some reason,
depended upon the fact that A0 C 0 lies above SQ, it is clear that the proof would not be
valid.
Of course you can go forward and back through the proof using the Next and Prev
buttons (or the n and p keys on the keyboard), and get a very good view of how the
proof works, including in situations with different configurations of the original triangle
4ABC.
Any of the diagrams in the book that are identified with a [P] after the file name
in the caption are proofs (or constructions) whose steps can be followed in the same
way with the Next and Prev buttons.

2.7 Making Measurements

Many geometry problems consist of a diagram with a few of the lengths and/or angles
specified and require the calculation of others. Computer geometry programs usually
allow you to assign various lengths to lines, et cetera, and then can display the lengths
of other lines. As a trivial example, if you construct a right triangle with legs of lengths
3 and 4 and ask to have the length of the hypotenuse displayed, you will be shown a 5
(or, thanks to computational error, perhaps 5.0000001, but that is at least a pretty good
hint that the true answer is 5).

2.7. MAKING MEASUREMENTS

27

This sort of problem is more typical of high school problem sets or mathematics
contests, and although Geometer can do this, it was primarily designed to work with
more general situations, so measurement manipulations may be a bit clumsy. Most of
the commercial programs are aimed specifically at high school students and hence may
be easier to use in this manner.

2.7.1 Finding an Unknown Length

The most obvious usage is for problems like the
one shown in figure 2.10. Given the measurements and the right angles in the drawing, find
the unknown length x.
In Geometer it is easy to draw a figure to
scale. Then if we divide the length of the segment labeled x by the length of the segment labeled 2 and multiply that result by 2, we obtain 10, the length of the segment x. This one
is pretty easy to do without a program like Geometer, but there are plenty of problems that
look this easy but are fiendishly difficult.

3
E

A2

D
x

Figure 2.10: An Unknown Length

Computer/Measure.T [D]

Even when Geometer says that x = 10.000 we do not have a solution; it might
be that x = 10.0001. But suspecting that x = 10 can lead to a quick solution: The
long sides are AB = 10 + 2 = 12 and AC = 5 + 3 = 8. Since 12 : 3 = 8 : 2
maybe the triangles are similar: 4AEB 4ADC. They obviously are since both
are right triangles sharing the angle at A, so knowing the answer (or, more accurately,
strongly suspecting that you know the answer) makes it trivial to prove that the answer
is correct.

2.7.2 Finding an Unknown Relationship

In figure 2.11, two circles
with centers at A and B are
tangent to a line. If a third
circle, centered at C, is tangent to both and to the line
as shown in the figure, what
is the relationship between
the radii of the three circles?
The Geometer diagram
in figure 2.11 is to help
gather data. In the diagram,
you can set the values of the
radii of the circles centered
at A and B, and the three

C
C

Figure 2.11: Mutually Tangent Circles

Computer/TanCircs.T [D]

28

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

radii are printed on the screen. (To change the radii, use Edit Geometry and change
them with the text editor.) For the diagram in figure 2.11, the values AA 0 = 0.500,
BB 0 = 0.500 and CC 0 = 0.125 are printed on the screen together with the diagram.
Actually, the Geometer diagram prints the values of both CC 0 and 1000CC 0 so you
can see more decimal placesGeometer normally shows only 3 places of precision
when it displays floating point numbers.
It is possible to make a diagram like the one above where you can set the radii with
the mouse, but due to the fact that a mouse can only select to the nearest pixel, it is
impossible in most cases to obtain nice round values for the inputs.
By repeatedly modifying the initial values for the radii, it is possible to obtain a
table of values like this:
AA0
1.000000
1.000000
1.000000
0.250000
0.111111

BB 0
1.000000
0.250000
0.111111
0.111111
0.111111

CC 0
0.250000
0.111111
0.062500
0.040000
0.027778

You can make this table as large as you want, but we have cheated a bit, selecting
values that may make the relationship a bit more obvious.
From the table above, it may be a lot easier to infer that the relationship is this: if
AA0 = 1/m2 and BB 0 = 1/n2 , then CC 0 = 1/(m+n)2 . Compare this with the Farey
circles in Section 2.11.2. (Remember that the values in that section are diameters, not
radii.)
 Actually, there generally are two circles tangent to the given circles and the line,
but will lie outside both circles. The only time this will not occur is if the original
circles are equal in size. If the radii of the original circles are 1/m2 and 1/n2 the outer
tangent circle will have radius 1/(mn)2. Note that if the two circles have equal radii,
then the outer tangent circle essentially becomes a line which is much like a circle
with infinite radius. If m = n, then 1/(m n)2 is very much like infinity.

2.7.3 Technical Details

Given the two radii as inputs, it is easy, using the pythagorean theorem, to determine
the separations of the centers of
the circles along the line. If the radii are r and R,
the separation of the centers is 2 rR which is easily constructed or calculated in a
computer geometry program.
The construction of the circle that is tangent to the two circles and to the line is a bit
more difficult. The easiest way to do it is to invert the two circles and the line through a
circle that is centered at the point of tangency of the two circles (see Chapter 8). Since
both circles pass through the center of the point of inversion, they will be inverted to
two parallel lines and the line will be inverted to a circle between them that is tangent
to both. From here, it is easy to find another circle of the same radius that is tangent to
the inverse of the line and to the two parallel lines which, when re-inverted, becomes
the circle tangent to all three.

2.7. MAKING MEASUREMENTS

29

2.7.4 Checking a Calculation

In Section 2.7.2 we used Geometer to find

a mathematical relationship between measurements. Sometimes it can be used in the opposite
way, to check an algebraic calculation.

In figure 2.12 we see a large circle with three

mutually-tangent interior circles that look a bit
like Mickey Mouse. If the radius of the surrounding circle is 1, the radius of the lower inscribed circle is R and the two equal circles
above have radius r, find the mathematical relationship between R and r.

O1 R

Figure 2.12: Tangent Circles

Computer/MMouse.T [M]

If x is the length of AO, we know that

AC = r, CO = 1 r, CO1 = R + r, and
AO1 = x + 1 R. The pythagorean theorem can be applied to triangles 4AOC and
4AO1 C to obtain two equations that relate x, r and R. A fair amount of tricky algebra
finally yields the following relationship between R and r:
r=

4R(1 R)
.
(1 + R)2

(2.1)

A couple of obvious values can be checked: R = 0 and R = 1 and these seem to

yield the correct value of r, but it is difficult to be certain of the calculation.
One way to check is to draw the figure and then use Geometers ability to calculate
and present lengths and to display both the value of r and of 4R(1 R)/(1 + R) 2 . As
the figure is modified, you can see immediately if the two values differ. In this case,
they remain the same, so you can be much more confident of the calculation.

2.7.5 Technical Details

It is a bit tricky to construct a Geometer diagram where the radius R can be modified and the two equal mutually-tangent circles are drawn. The easiest method involves inversion in a circle that is discussed in Chapter 8. In addition, you will need
to know how to do calculations in Geometer, and that is discussed in Section 11.5. To
see the details, examine the text of the Geometer file Computer/MMouse1.T. (The file
Computer/MMouse.T is simply for making the illustration in the text; it does not contain
the Geometer code that does the calculation.)
Another alternative is to use the result of your algebraic calculation in equation 2.1
to draw the two circles. The centers of those circles must lie on a circle of radius 1 r
from the center of the surrounding circle of radius 1 and must also lie on lines parallel
to AO and at distances of r from it. If the calculation is correct, two circles of radius
r drawn about these centers should appear to remain tangent to each other and to the
other two circles as the radius R is modified. You can modify R and visually check

30

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

the tangencies. You can examine the text of a Geometer file that uses this technique in

Computer/MMouse2.T.

One other minor technical point that is illustrated in the file Computer/MMouse1.T
is that in a complicated diagram it is sometimes hard to identify the points needed to
do a calculation. Which are the points, after all, that determine the radius of the various
circles? Using the graphical user interface, simply color those points differently and
then it is easy to see in the text editor which points are of interest. For example, in this
diagram, the points at the center and edge of the circle of radius R are in red, so in the
text editor we merely need to look for points with the text .red in their description.

2.8 Searching for Patterns

The Geometer program has a mode where
B
the program searches for patterns that may
hold in a drawing. After a diagram is drawn,
D
Geometer can be instructed to look for patG E
terns, at which point it searches for unusual
A
situations in the diagram such as three points
F
C
that lie on a line, or four points on a circle,
or angles that are equal or complementary or
supplementary, et cetera. Then as you manipFigure 2.13: Searching for a Proof of
ulate the diagram, Geometer rechecks these
Miquels Theorem
relationships over and over to see if they conComputer/Pattern.T [M]
tinue to hold. When you tell Geometer to
stop searching for patterns, it will tell you about any relationships that always held.
Obviously these are not necessarily always true, and they are true only to the degree
of accuracy to which Geometer can do arithmetic calculations.
But knowing, for example, that four seemingly unrelated points always lie on a
circle is often tremendously important in finding a proof of whatever theorem you are
trying to prove.
As a simple example, let us consider again Miquels theorem that we examined in
Section 2.2.1. This time, instead of simply convincing ourselves that the theorem is
true, we will try to find a proof.
Begin with the same diagram that we used in Section 2.2.1 and we will add a point
G that is the intersection of circles BDE and AF D. Connect G to all six other labeled
points, and then ask Geometer to search for patterns using the Test Diagram and End
Test commands. The diagram should look something like what is shown in figure 2.13
Move points A, B, and C a bit, and then end the test. A small auxiliary window
appears on your screen that contains the following information:
Points on a circle:
C E F G
Equal angles:

2.8. SEARCHING FOR PATTERNS

(A D G) (B E G)
Supplementary angles:
(A D G) (B E G)
are supplementary to
(A F G)
Supplementary angles:
(B C A)
is supplementary to
Supplementary angles:
(B A C)
is supplementary to
Supplementary angles:
(A B C)
is supplementary to
Concurrent circles:
c1 c2 c3

31

(E G F)
(D G F)
(D G E)

First of all, it tells you that the point G is also on the third circle (CEF G) which
you are trying to prove, and it also tells you that the three circles are in fact concurrent.
(Geometer uses the internal names c1, c2, and c3 of the circles since they have no
user-defined names1 . The points are all listed using the names you gave them.)
But there is also information about equal and supplementary angles. Just look at
one: BCA is supplementary to EGF . This is the key to the proof! You know that
two sets of four points: BEGD and AF GD are concyclic, so F AD is supplementary
to DGF and DBE is supplementary to EGD. The three angles of 4ABC add
to 180 so with a little arithmetic, it is easy to show that EGF = 360 DGF
EGD is supplementary to ACB, which is enough to complete the proof.
The lines connecting G to A, B, and C were unused, but since we had no idea how
to start the proof, it did not hurt to test them. For all we knew, the construction of those
lines might have been the route to the solution.
Geometer attempts to show only interesting relationships. In the example above,
it only showed one set of four points lying on a circle when in fact there are three. But
the first two sets were required by the constraints on the diagram to lie on a circle.
Similarly, if a line is defined by two points and you constrain a third point to lie on that
line, Geometer will not tell you that those three points are on a linethey had to be
on a line.
Searching for patterns is a very powerful proof finding technique. Many of the
relationships displayed by Geometer will be the entries on the left side of a highschool statementsreasons two-column proof. You just have to put them in the
correct order.
Although it usually works, the technique is not foolproof. The relationships are
only calculated to an accuracy limited by the floating point precision of the machine,
so as far as the internals of the program are concerned, the relationships hardly ever
hold exactly. But if they hold to a certain tolerance and continue to hold as the diagram
is modified, Geometer will report them at the end.
1 By default, only points have names, but you can, if you wish, assign names to other Geometer primitives.

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

32

2.9 Searching for Extremes

This is just another measurement, in a sense, but suppose you have a question like this:
What is the minimum perimeter of a triangle whose vertices lie on the edges of a given
triangle? You can draw the triangle constrained to have its vertices on the edges of
the given triangle and display its perimeter. Then you can manipulate the inner triangle
as you watch the value of the perimeter and easily find a very close approximation of
the triangle that does the trick. If the original triangle has all acute angles, the answer
is the triangle whose vertices are the bases of the altitudes of the original, which you
may be able to notice by inspection. What happens if the original triangle contains an
obtuse angle? (See Fagnanos problem in Chapter 12.)
A different, but similar, example is described below.

2.9.1 Maximum Area

Suppose that a pentagon is inscribed in a circle. What is the arrangement of the vertices that gives the maximum area? It is easy to create a Geometer file that lets you
manipulate the vertices and continuously prints the area. With a random arrangement
of points on the left in figure 2.14, a certain area is enclosed.
You can then slide any of the points on the circle, constantly trying to find larger
areas, and after some playing around, you will come up with a picture like that on the
right in figure 2.14 where it seems impossible to increase the area. Notice that the
pentagon in this second figure appears to be very nearly a regular pentagon with all
sides and angles equal.
In fact, the regular pentagon is the largest possible; this exercise does not prove
that to be the case, but it certainly provides plenty of visual evidence which would
encourage you to seek a rigorous mathematical proof. An interesting related exercise
is to determine the size of the largest equilateral triangle that can be inscribed in a
square.
Another interesting example can be examined in the file Computer/MaxQuad.T that
illustrates the fact that given the lengths of the sides of a quadrilateral, the quadrilateral
of maximum area that can be produced from those sides is the one that is concyclic.

2.9.2 Technical Details

Geometer has a command available through the text editor to evaluate the area of a
polygon. The code looks like this:
poly1 = .polygon(5, v6, v7, v3, v4, v5);
area = .f.area(poly1, "Area");

The first line above was generated by Geometer from the graphical user interface
by clicking on P..P=>Poly and then clicking on the points around the pentagon in

2.10. LOCUS OF POINTS

Area = 1.314

33

Area = 1.521
B

D
D
A
E

Figure 2.14: Area of a Pentagon

Computer/PolyArea.T [M]

order and then clicking on the first point to close and complete the polygon. The
internal name is poly1, so the .f.area command takes poly1 as a parameter, and a
floating point number called area is generated. Since the object called area contains
the text "Area" in its definition, the value of the float is continuously displayed on the
Geometer screen, something like this: Area = 1.132.

2.10 Locus of Points

If you are asked to find the set of points equidistant from a fixed point and a line,
it is trivial with Geometer to construct one such point. If that point is colored in a
smearing color, and is dragged around, you will discover that it paints hundreds of
points, all of which seem to lie on a parabola. Although this is not a proof that the
locus is a parabola, it provides almost overwhelming evidence. This can, of course, be
done by hand, but the process of doing so is much more painful and error-prone. In
addition, once the diagram that illustrates this relationship is drawn, you can move the
fixed point relative to the fixed line and see how the resulting parabola changes size
and shape.

2.10.1 Parabola as a Locus of Points

Here is how to use Geometer to draw the locus described above. First, draw a line L,
a point P not on L, and constrain a point X to lie on L. All points equidistant from X
and P lie on the perpendicular bisector of those points, and the point of interest is on
the line through X and perpendicular to L. The intersection of those lines is a point
equidistant from X on L and the point P . As X is moved along L, all such points are
drawn in a smearing color which generates the diagram in figure 2.15. Note that you
can move P relative to L (before sliding X back and forth) to generate different shapes
of parabolas.

34

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

Figure 2.15: Parabola as a Locus of Points

Computer/Parabola.T [M]

2.10.2 Cardioid/Limacon
It is almost equally easy to generate a far
more interesting family of curves including
the cardioid and the limacon of Pascal. Figure 2.16 shows a limacon since the point P
does not lie on the circle.

Given a circle K and a point P not on the

circle, the limacon of Pascal is the set of all
points that are the projection of P on the lines
M
tangent to the circle K. In figure 2.16, you
can move the point M around the circle and
trace out a limacon. If the point P happens
Figure 2.16: The Limacon of Pascal
to lie on the circle, a cardioid is generated inComputer/Limacon.T [M]
stead. It is interesting to explore the shapes
generated when the point P is different distances from the circle, including on and inside the circle K.

2.11 Computer-Generated Diagrams

In Geometer, every diagram is saved as a file in text format that can be inspected and
edited. Geometer itself provides a primitive editor, but there is no reason you cannot
use your favorite powerful text editor to do the same thing. In addition, you can write
a computer program that generates a very complicated diagram simply by generating
the text form of a Geometer file.
Geometer is usually used with the standard graphical user interface, but at times,
it it easier to use the built-in text editor to make some changes. One of the beauties of

2.11. COMPUTER-GENERATED DIAGRAMS

35

the program is that you can flip back and forth: use the graphical interface when it is
easy, switch to the text editor for a while, then back to the graphical interface, et cetera.
This is an advanced technique, but it can be incredibly powerful.
Warning: Unless you understand the internal form of the Geometer diagram files,
some of the examples in this section may be a bit mysterious. The examples are not
difficult, however, and anyone familiar with computer programming should have no
difficulty understanding what is going on.
Some diagrams are difficult to generate because they are large, or because they
require repeated complex operations, or they require lots of specific exact coordinates.

2.11.1 Drawing a Grid

As a very simple example, suppose you
want to draw a grid of equally-spaced
lines with 51 lines in each direction to
make a grid of 5050 squares as in figure 2.17. It is quite difficult to generate
something like this by hand, but with a
tiny computer program written in your
favorite language (the authors favorite
language is C or C++), the generation
becomes trivial.
What you need is simply a file
that looks like 51 copies of the first
6 lines of the Geometer code below,
but where each successive set of lines
has different sets of endpoints for the
lines. The second 6 lines illustrate
what should appear as the second set
of lines. The pattern continues for 51
sets of those six lines.
v01
v11
lh1
v21
v31
lv1
v02
v12
lh2
v22
v32
lv2

=
=
=
=
=
=
=
=
=
=
=
=

Figure 2.17: A 51 by 51 Grid

Computer/Grid.T [D]

.pinned(-1, -1.000000, .in);

.pinned(1, -1.000000, .in);
.l.vv(v01, v11);
.pinned(-1.000000, -1, .in);
.pinned(-1.000000, 1, .in);
.l.vv(v21, v31);
.pinned(-1, -0.960000, .in);
.pinned(1, -0.960000, .in);
.l.vv(v02, v12);
.pinned(-0.960000, -1, .in);
.pinned(-0.960000, 1, .in);
.l.vv(v22, v32);

Here is a tiny program written in the C language that will generate a Geometer file
that will do the trick:

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

36
#include <stdio.h>
#define GRIDSIZE 51
main() {
int i;
double x;
int n = 1;

for(i = 0; i < GRIDSIZE; i++) {

x = -1.0 + i*2.0/(GRIDSIZE-1.0);
printf("v0%d = .pinned(-1, %f, .in);\n", n, x);
printf("v1%d = .pinned(1, %f, .in);\n", n, x);
printf("lh%d = .l.vv(v0%d, v1%d);\n", n, n, n);
printf("v2%d = .pinned(%f, -1, .in);\n", n, x);
printf("v3%d = .pinned(%f, 1, .in);\n", n, x);
printf("lv%d = .l.vv(v2%d, v3%d);\n", n, n, n);
n++;
}
}

The 306-line file generated by the program above produces figure 2.17.

2.11.2 Farey Circles

The example above is quite simple; here is one that is more interestingit displays
some of the so-called Farey circles.
The Farey circles consist of all circles of diameter 1/q 2 that are tangent to the real
number line at the point p/q, where p/q is a rational number reduced to lowest terms.
See figure 2.18. The amazing thing about these circles is that none of them pass through
another, but many pairs are tangent.

Figure 2.18: Some Farey Circles

Computer/farey.T [D]

To generate a list of circles for Geometer to draw, we would like to make a sequence of entries in a Geometer file that look like this:
v6 = .pinned(2.500000, 0.125000, .in);
l6 = .l.vlperp(v6, l, .in);
vv6 = .v.ll(l, l6, .in);
c6 = .c.vv(v6, vv6);

2.11. COMPUTER-GENERATED DIAGRAMS

37

In the example above, the circle above the point 2.5 on the real line corresponds to
the rational number 5/2, or p = 5 and q = 2. This circle must be centered at (2.5, .125)
since its diameter is 1/q 2 and hence must have radius 1/8 = .125. To construct this
circle in the example code above, the point v6 is pinned with coordinates (2.5, 0.125).
From that point, a perpendicular line l6 is dropped to the line l, which is defined once
in the file as the real axis. The point vv6 is the intersection of those lines, and finally,
the circle that interests us is centered at v6 and passes through vv6.
Here is the C code that generates the complete Geometer file, including code to
generate the line l:
#include <stdio.h>
#define NMAX 12
int gcd(int i, int j) {
if (i % j == 0) return j;
return gcd(j, i%j);
}
void outcircle(int q, int p) {
static int n = 0;
printf("v%d = .pinned(%f, %f, .in);\n", n,
((float) p)/((float) q), 0.5/((float)q*q));
printf("l%d = .l.vlperp(v%d, l, .in);\n", n, n);
printf("vv%d = .v.ll(l, l%d, .in);\n", n, n);
printf("c%d = .c.vv(v%d, vv%d);\n", n, n, n);
n++;
}
main() {
int p, q;
printf("v00 = .pinned(0, 0, .in);\n");
printf("v01 = .pinned(1, 0, .in);\n");
printf("l = .l.vv(v00, v01, .in);\n");
outcircle(1, 0);
for (q = 1; q <= NMAX; q++)
for (p = 1; p <= 3*q; p++)
if (gcd(q,p) == 1) outcircle(q,p);
}

The constant NMAX is 12, which is the largest denominator that the program will
consider. The routine gcd finds the greatest common divisor of two numbers which
is used to make certain that the fractions are reduced to lowest terms. The routine
outcircle prints all the data needed to draw a circle over the point p/q with radius
1/q 2 . The main routine prints Geometer code to draw the line l at y = 0 and then
marches through all possible combinations of q and p and generates circles when p/q
is a fraction reduced to lowest terms. This C code with NMAX equal to 12 generates a
file containing 559 lines of Geometer code.
Here is an amazing property of the Farey circles that you may wish to investigate.
If the two circles corresponding to the fractions a/b and c/d are tangent to each other
and to the line y = 0, then the circle corresponding to (a + c)/(b + d) is tangent to
both of them as well as to the line y = 0. See [Conway and Guy, 1996].

38

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

Note: the radii of the Farey circles can be calculated using the Descartes circle
theorem (see Section 2.14) where we consider the straight line to be a circle of infinite
radius (or equivalently, zero curvature).

2.12 Animations
This is again a useful feature for teachers of geometry. The Geometer diagrams can
be programmed to run through a fixed sequence of steps when the user presses the Run
Script button. An animation of a process often makes things much easier for the students to visualize. This is also an advanced technique that requires some programming
in the Geometer diagram file.
The spirograph example in Section 2.15.3 is really an animation where the angles
in the virtual spirograph disks are incremented in a series of successive equal steps, but
a far more interesting one is illustrated in the series of images in figure 2.19 that can be
used to convince a student that the area of a circle of radius r is given by the formula
r2 .
The animation divides a circle into a number of wedges, moves the wedges into two
lines, and then pushes the lines together to form a rectangle. The three figures illustrate
successive steps in this process which is continuous on the computer screen.

Figure 2.19: Circle in Wedges

Computer/Circle.T [S]

2.12.1 Technical Details

This particular demo is perhaps the most complex that has ever been written for Geometer. For each of the wedges, a computer program written in C generated a translation/rotation pair that would gradually move it from its initial position on the circle into
the final alignment. A script was written that steps a variable from 0 to 1, and while the
variable is less than 0.5, the translation/rotation is performed. While it is larger than
0.5, a translation from the final unwrapped position is performed to push the two
halves together. The coordinates for the slices were also computer generated.

2.13. DRAWING FOR PUBLICATION

39

2.13 Drawing for Publication

Geometer is a great drawing program, and in fact, it was used to generate every illustration in this book. When you get a diagram that you like on the screen, Geometer can produce from that diagram an EPS (Encapsulated PostScript) file that will
draw the same figure at any magnification, and in a form that can be included in many
text manipulation programs. Painting and image manipulation programs like Adobes
Photoshop can convert EPS files to almost any other graphics file format. All of
Chapter 11 is devoted to making drawings and presentations easier to understand.
One advantage of a computer geometry
program for producing illustrations that usually becomes obvious only when you try to
do it is that with a program like Geometer,
after drawing the figure, you can adjust the
size and shape to fit the space available on
the page, you can adjust the shape so that important labels do not fall on top of each other,
et cetera.

2
6

7
5

Here, the word publication means anything where you would like to draw an accurate figure, from a published book to a handFigure 2.20: Dirichlet Domains
out for class. All the figures in this book were
Computer/Dirichlet.T [M]
produced using Geometer, but the example
in this section was originally prepared for a talk on computational geometry, not Euclidean geometry.

2
6

7
5

A set of n distinct points in the plane divide the plane into n distinct regions called
Dirichlet domains where the points in each
region are closest to one of the points. Points
on the boundaries of the regions are equidistant from two or more points. Figure 2.20
illustrates the Dirichlet domains for a set of 7
points.

It is not difficult to generate the domains.

All the boundaries between the domains are
segments or rays of the lines that are perpenFigure 2.21: Dirichlet Domain Con- dicular bisectors of the segments between the
points. Some of those segments connecting
struction
pairs of points are displayed in figure 2.21
Computer/Dirichlet.T [M]
and it is obvious that the boundaries of the
Dirichlet domains are perpendicular to those segments.
To construct the diagram, the perpendicular bisectors were drawn, their key intersections were identified with additional points, and then the lines themselves were
made invisible. Next, visible lines were drawn only along the true boundaries of the

40

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

domains. Notice that not all the perpendicular bisectors are used; the ones used depend
on the locations of the points.
Since this diagram was solely for publication, and not for manipulation, it will
work only if the points determining the domains are not moved too much. To see what
happens with too much movement, load the Geometer diagram and try making both
small and large movements of some of the points. The diagram is slightly flexible, but
with too much variation, it breaks down. Since it was only to be shown in printed form,
that was not a problem.

2.13.1 Technical Details

To draw the Dirichlet domains, first the set of seven points was placed in reasonable
positions. Then all the perpendicular bisectors were drawn as lines that extend forever
in both directions in an unusual color. By examining the particular configuration of the
points, it was easy to see by inspection which sections of the long lines mattered in the
final figure.
Points were placed at the key intersections, and those were connected by line segments of the final color. Once the key part of a perpendicular bisector was redrawn as
a segment in the correct color, the infinite bisector was changed to the invisible color.
This process was repeated until the entire Dirichlet domain illustration was complete.
Obviously, this only works for small movements of the points; larger movements
will generate totally bogus results.

2.14 Computer Algebra Systems

Although we will do almost nothing with them in this book, there are commercial
computer algebra systems such as Mathematica, Maple and Macsyma that are
designed to manipulate algebraic expressions. Since it is often easy to write down a
set of equations that relate various geometric measurements but very difficult to solve
them, computer algebra systems can provide a powerful tool to manipulate and solve
these equations.
As an example, consider the configuration of circles illustrated in figure 2.22. Given
three pairwise-tangent circles, there are two other circles that are tangent to all three.
In the figure, the one enclosed by the three is shown, and there is an additional circle
(not shown) that surrounds the three circles. Circles in this configuration are sometimes
called Soddy circles.
A relationship between the curvatures of the circles is given by the circle theorem
of Descartes. The curvature of a circle is the inverse of the radius, so if the radii of
the four circles are ra , rb , rc and rd , then their curvatures are a = 1/ra , b = 1/rb ,
c = 1/rc and d = 1/rd which are related by the following formula:
2(a2 + b2 + c2 + d2 ) = (a + b + c + d)2 .

2.14. COMPUTER ALGEBRA SYSTEMS

In the figure, we can see
that four triangles are formed
the outer one, 4ABC, and it is
composed of three inner triangles.
A similar diagram can be drawn
would relate the radii with the surrounding tangent circle. The sides
of the outer triangle have lengths
ra + rb , rb + rc and rc + ra and a
substitution of those lengths into
Herons formula (see Section 5.6)
tells us that the area A(4ABC)
is given by:
p
ra rb rc (ra + rb + rc ).

41

B
rb
rb

rb

rc
rd
rd
ra
A

rd

rc

rc
ra
ra

Exactly similar expressions

exist for the three triangles that
Figure 2.22: The Descartes Circle Theorem
compose it and their areas add to
Computer/Soddy.T [M]
give the area of the larger triangle
so the radii satisfy the following
equation:
p
p
p
ra rb rc (ra + rb + rc ) =
ra rb rd (ra + rb + rd ) + ra rc rd (ra + rc + rd )
p
+ rb rc rd (rb + rc + rd ).

It is conceptually easy to eliminate the radicals from the equation above. If we let
A, B,C andD represent
the expressions under the radicals, the equation looks like

this: A = B + C + D which can be simplified as follows:

A =
B+ C+ D

A B =
C+ D

( A B)2 = ( C + D)2

A + B 2 AB = C + D + 2 CD

A + B C D = 2 AB + 2 CD

(A + B C D)2 = (2 AB + 2 CD)2

(A + B C D)2 = 4AB + 4CD + 8 ABCD

(A + B C D)2 4AB 4CD = 8 ABCD

((A + B C D)2 4AB 4CD)2 = 64ABCD
This seems simple enough, but recall that A, B, C and D are each complex expressions. When the final equation above is expanded, there are approximately 400 terms
of degree 16. This can be factored and substitutions made of 1/a for r a , et cetera, and
Descartes theorem drops out. When printed, the complete Maple calculation requires

42

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

about 18 pages of equations. This is not something that you would like to do by hand,
but it only requires a few minutes on a computer.
Note: Descartes circle theorem is quadratic so for every set of three radii (or
curvatures) there are two solutions for the fourth. If the solution is negative, the result
is the surrounding circle; if positive, it is the surrounded circle.

2.15 Non-Mathematical Uses of Geometer

We can often use Geometer as a general drawing program where the ability to satisfy
geometric constraints is simply a useful feature. It has been used to illustrate mechanical linkages, to draw illustrations from computational geometry and chemistry,
or simply to produce works of art. In fact, Appendix B contains a description of the
somewhat artistic illustrations that appear at the beginning of each chapter of this book.
This section includes three uses for the
Geometer program that are a bit outside the
realm of geometry, one in mechanical engiC
neering, one in optics, and one that simply
O
l
generates a pretty picture.
A
A
r

2.15.1 Peaucelliers Linkage

For many years, it was unknown whether a

mechanical linkage could be designed that
Figure 2.23: Peaucelliers Linkage
would convert circular motion (such as the
Computer/Peaucellier.T [P]
output of an engine) into perfectly linear
motion. Peaucelliers linkage does just that,
as shown in figure 2.23. (The diagram is
marked with a [P] because not only does it allow you to manipulate the linkage, but
it contains a Geometer proof that the linkage does, in fact, generate a straight line.)
In the figure, the point O is fixed on the circle and the point A is constrained to
move around the circle. The segments OC = OD and AC = CA0 = A0 D = DA
are solid bars that are connected together with pivots at O, A, C, A 0 , and D. You can
change the lengths of those bars by moving the endpoints of the segments labeled l
and r on the left side of the figure.
As you move the point A around the circle, the point A0 traces out a straight line, a
few points of which are shown in figure 2.23.
It is interesting to construct a similar figure where the point O does not lie on the
circle. What is the relationship then between points A and A0 ?

2.15.2 Focus for a Spherical Lens

Snells law in physics allows you to calculate the angle through which light is bent
when it passes a surface between two substances with different indices of refraction. In

2.15. NON-MATHEMATICAL USES OF GEOMETER

43

fact, the index of refraction is just the ratio of the sines of the angles of incidence and
refraction.

L
Figure 2.24: Good Focus with a Spherical Lens
Computer/lens.T [M]

P
O

L
Figure 2.25: Bad Focus with a Spherical Lens
Computer/lens.T [M]

In this example, we imagine a sphere of glass with index of refraction 1.5 and light
emitted from a point O outside the glass. When light from O strikes the glass at a point
P , the Geometer diagram finds the angle of incidence with the glass, calculates the
angle of refraction and draws a ray from P in that direction. As you move the point P
on the surface of the glass, all those rays are smeared onto the diagram.
Figure 2.24 shows that if P is only a little bit away from the axis of the lens, the
rays very nearly focus, but if there is more movement, as in figure 2.25, the rays do
not come close to being in focus. The shape of the lens and the position of O can
be changed in the diagram to experiment with different lens shapes and with different
object distances.

2.15.3 Pure Art: A Spirograph

As children, most of us played with a Spirographa toy consisting of geared plastic
circles that could be rotated around each other, and as they did, they dragged a pencil
that generated wonderful patterns. With the plastic version, you were pretty much
limited to two gears, but it is easy to write a short Geometer program that has any
number. The example shown in figure 2.26 consists of three virtual gears, and requires
only a 21 line Geometer file.

44

CHAPTER 2. COMPUTER-ASSISTED GEOMETRY

2.15.4 Technical Details

The lens demonstration began with
standard Geometer commands to
draw and control the positions and
shape of the lens and the position of
the object point to be focused. Then
a point was constrained to be on the
surface of the lens, and a line drawn
from there to the object. A line from
the center of the circle was drawn to
the point on the surface (which would
be perpendicular to the lens surface).
This and the incoming line segment
form an angle whose sine is calculated using Geometers internal calculator and multiplied by an appropriate constant to find the sine of the
Figure 2.26: A Fancy Spirograph Drawing
angle of refraction. A new line in the
Computer/Spirograph.T [S]
smearing color was generated passing
in that direction, and all other construction aids were changed to the invisible color.
The spirograph points are simply calculated as a series of three rotation/translation
pairs, as if three gears with different radii and hence numbers of teeth were hooked one
to the other in a sort of super-spirograph. Examine the Geometer code to see exactly
how this is done. Using a script that takes all the gears through one complete cycle,
a single point at the origin is put through the three translation/rotation transformations
and a point is drawn in the smearing color at the final resulting location.

Chapter

Geometric Construction
To use Geometer or any computer geometry program to investigate new problems you
will need to be able to draw diagrams that correspond to those problems. In principle,
a vast collection of drawings can be made with only two toolsa straightedge and a
compassbut to draw complex diagrams quickly, the ability to use a larger set of tools
is extremely valuable.
The classical Greek construction problems that allow only the straightedge and
compass as tools are certainly interesting in their own right, and there is a large literature associated with them. In this chapter when we wish to refer to that sort of problem,
we will call it a classical construction problem.
We will examine the classical construction problems later in Section 3.11, but if
you are planning to use a computer geometry program, you are going to be forced to
face your own construction problems right away.
If you would like to illustrate a theorem with a computer drawing that can be manipulated, you need to draw it so that it has the same degrees of freedom that the theorem
45

46

CHAPTER 3. GEOMETRIC CONSTRUCTION

allows. Luckily, computer arithmetic, although not perfect, is accurate enough that
your drawings will seldom be in error by more than a pixel or so, if they are drawn
correctly.
Every computer geometry program, including Geometer, certainly supports the
ability to do classical straightedge and compass constructions, but all of them have
additional capabilities. Some of these capabilities are just shortcuts to skip some steps
in a classical construction, but some allow you to do things that are far more powerful
than what a straightedge and compass would allow (see Section 3.12).
Of course all the examples here are based on Geometers tools, although it is likely
that similar operations are available in other programs.

3.1 The Circumcenter

As a warm up, suppose you want to examine the circumcenter of a triangle and to
see how it moves as the size and shape of the triangle changes. (The circumcenter is
the center of the circle passing through the three vertices of the triangle.) Your goal,
therefore, is to make a Geometer diagram with a triangle 4ABC, all of whose vertices
are movable, and with the circumcenter of that triangle visible and constructed in such
a way that it adjusts its position automatically as the vertices of the triangle are moved.
The construction can be perG
formed with classical techniques.
B
The center of the circumcircle
must be equidistant from the
E
three vertices, so one way to conO
C
D
struct it is to find the perpendicuF
lar bisectors of the segments AB
A
and AC (or any other pair of the
three segments) and their intersection is the circumcenter. All
Figure 3.1: Construct the Circumcenter
the points on a perpendicular biConstruction/Circumcenter.T [M]
sector of a segment are equidistant from the endpoints, so the point at the intersection of the bisectors will be equidistant from all three vertices. This will be the required center of the circumcircle.
The usual classical construction of the perpendicular bisector of a segment involves
drawing circles centered at each endpoint and passing through the other endpoint. The
two intersections of those two circles determine a line. This line is the perpendicular bisector of the segment. The entire classical construction of the circumcenter is
illustrated in figure 3.1.
If you are beginning to use Geometer, this construction is a good exercise. You
will begin to learn where the controls are, and you will see how a construction can be
carried out with a minimum number of tools. As you are drawing, especially since
you may have to stop and think about where to find the tools you need, you may get
confused about which parts of the construction are which. Remember that you can

3.2. AVAILABLE TOOLS

47

change the colors of lines, points and circles, and you can use that sort of marking to
help remind you where you are in the construction. For example, color the lines of
the original triangle differently, and use red construction lines to get the perpendicular
bisector of the red side and green ones when you are working on the green side.
The classical approach requires eleven steps after the triangle is available. Four
circles are drawn, four intersections of the circles are determined, two lines are drawn
through these intersection points, and finally the intersection of those lines is the required circumcenter.
This is a perfectly good construction, but as you can see, there are a lot of auxiliary
lines, and altogether it required eleven steps to find the circumcenter.
Geometer has shortcuts. For example, there is a built-in command to construct the
perpendicular bisector, after which only their intersection would need to be found for a
total of three steps. But there is an even faster way: Geometer has a command (called
PPP=>C in the command area under the Circles label) to construct a circle passing
through three points (this would be the circumcircle) and another to find a point at
the center of a circle, so the lazy Geometer user can construct the circumcenter of a
triangle in two steps. The command to find the point at the center of a circle is in the
Primitives pulldown menu under the Point submenu and is called C=>P Ctr.

3.2 Available Tools

Even if you do not know how to use all of them, it is good to have at least a vague
idea of what tools are available in your computer geometry program. Then if you are
having trouble with a construction, you may decide that it would be a good idea to learn
exactly how a couple of likely commands might be used.
In Geometer, the more common construction tools are listed in the menu to the
right of the drawing area, organized by the type of primitive they createat the top,
the point and line creation commands, and so on. This list is not complete, however.
In the Primitives pull-down menu are a lot of others, all of which are available via the
graphical user interface.
If you are willing to write Geometer textual descriptions directly, there are still
more commands available, but most people will not need to use these. The textual
descriptions include quite powerful calculation tools that will allow you to draw almost
anything.

3.3 The Structure of a Geometer Diagram

Think of a Geometer diagram as a list of commands, each of which is executed exactly
once beginning with the first and continuing to the end of the list. There is no looping

48

CHAPTER 3. GEOMETRIC CONSTRUCTION

and branching can occur only in a very limited sense. The vast majority of diagrams
consist of a linear list of commands that are issued one after the other 1 .
When you view a diagram on your computer screen, you can only move points, and
only certain free points at that. There are completely free points that you can move
anywhere, and there are partially constrained points that you can move, but they must
remain on a line or a circle or a conic section. Other than these free points, everything
else is constrained by other geometric objects.
A line that is constrained to pass through two points will change if you move either
of the two points upon which it depends. Those points can change either because they
were free and you moved them directly with a mouse, or because they were constrained
by other objects that moved. For example, if a line is constrained to pass through a free
point and you move that point, the line will move, and so will the point defined to be the
intersection of that line with a circle. So if another line passes through that line-circle
intersection, it too will move, and so on.
Sometimes a constraint fails to hold. For example, if a point is defined to be at the
intersection of a line and a circle and the line moves so that it no longer intersects the
circle, the intersection point becomes undefined, and it, as well as any other objects
that depend on it, are simply not drawn.
Remember that there are hundreds of examples of Geometer diagrams on the CD
and in addition to examining any of them with the Geometer program, you can also
look at them with a text editoreither the one contained within Geometer itself or
with your own favorite. The Geometer diagrams are pure text files.

3.4  Interpreting Geometer Files

For complete details, see the Geometer reference manual on the CD, but we will examine here a small Geometer program and see how to interpret it. We will use as an
example the classical construction of the circumcenter discussed earlier and displayed
in figure 3.1, above. Here are the text file contents:
.geometry "version 0.60";
v1 = .free(-0.210169, -0.634407, "A");
v2 = .free(-0.39322, 0.484237, "B");
v3 = .free(0.183051, -0.18017, "C");
l1 = .l.vv(v1, v2);
l2 = .l.vv(v2, v3);
l3 = .l.vv(v3, v1);
c1 = .c.vv(v1, v2);
c2 = .c.vv(v2, v1);
v4 = .v.cc(c1, c2, 2, "D");
v5 = .v.cc(c1, c2, 1, "E");
c3 = .c.vv(v3, v2);
c4 = .c.vv(v2, v3);
1 In practice, Geometer does not need to execute all the commands every time. Many operations leave
most of the diagram unaltered, and Geometer can usually detect this and avoid a lot of extra calculation. But
the net effect is as if it had executed every instruction from beginning to end.

3.5. SIMPLE CLASSICAL EXAMPLES

v6
v7
l4
l5
v8

=
=
=
=
=

.v.cc(c3,
.v.cc(c3,
.l.vv(v4,
.l.vv(v7,
.v.ll(l4,

49

c4, 2, "F");
c4, 1, "G");
v5);
v6);
l5, "O");

The first line does nothing for the drawing but is used by Geometer for version
control. Files missing this line will work without it.
The next three lines define three completely free points whose internal names are
v1, v2 and v3. The .free signifies the type of object, a free point in this case. The
current x- and y-coordinates are given by the next two numbers, and the names in
quotation marks are the names that appear on the screen next to those points. If there is
no name in quotes, the point will appear without a name, although of course it retains
the internal Geometer name.
All the words in the file that begin with a period, like .free, .l.vv, et cetera, are
reserved Geometer commands; the others are variables whose names you or Geometer
can create.
The next three lines of code draw three lines: the first connecting v1 with v2 and
so on. The form of the command, .l.vv tells you a lot about what it means. The first
l means line, which is the type of object being created. The l is followed by
another period, and then a list of the types of objects it depends upon, in this case, the
line depends upon the positions of two points (called vertices in computer graphics
lingohence the vv). Similarly, .c.vv indicates a command that makes a circle
from two points (vertices)the center and a point on the edge, and .v.cc makes a
point at the intersection of the circles. Of course there are up to two intersections of a
pair of circles, so the parameter 1 or 2 tells Geometer which one to use.
You may find it very interesting to draw diagrams using the graphical user interface
and then look at the generated text form of the code. You can do this with the Edit
Geometry command in the Edit pulldown menu. Try changing colors of objects,
changing the line types, angle types, et cetera, and see how the code is affected. You
can also edit code while you are in the text editor, although it is quite easy, at first, to
mess things up pretty badly, and Geometer does not have a very robust error checker.
Note: If you are trying to edit the code and get completely messed up, the easiest
thing to do to get out of the mess is to select all the code in the file and delete it. Then
save the file, and Geometer is perfectly happy to read a completely empty text file and
continue.

3.5 Simple Classical Examples

The best way to learn to do constructions with Geometer or any similar program is to
examine a series of examples, beginning with easy ones and progressing to ones that
are more complex.
In this first section we will show how the most basic commands can be used to
build up common constructions: the midpoint of a segment, a perpendicular to a line
from a point, a tangent to a circle from an exterior point, and so on.

50

CHAPTER 3. GEOMETRIC CONSTRUCTION

Although almost all the examples in this section are available directly as Geometer
commands, all can be done as a series of smaller steps with only the most primitive
straightedge and compass operations.
Even if you plan to use Geometer for your work and the operations below are
available as primitive Geometer commands, it is useful to read through these examples
as they demonstrate how more complex operations can be built up from primitive ones.
In most computer geometry programs (including Geometer) you can define macros
which are combinations of simple commands that can be applied as a unit. Thus if
you compose a macro to find the common exterior tangent lines to a pair of circles,
you can use that macro over and over again, whenever you need to do that particular
construction as part of a more complex problem.
Note that in order not to clutter up the illustrations too much, we often draw only
an arc of the required circle in the sections that follow, because we only care where
the circle intersects some other line or circle. In every case, we could have drawn the
whole circle, but it is easier to follow the steps with less junk in the drawing.
In the first few examples, the Geometer commands that perform the given steps
are indicated in the text. For example, the command to construct a line connecting two
points is labeled as PP=>L on the Geometer control panel.

3.5.1 Bisect a Line Segment

Given a line segment AB, construct M ,
the midpoint of that segment.
If you use Geometer you can load
the file Construction/Segbisect.T to
follow the construction more easily on
your computer screen. The caption
under the figure tells you where to
find the file: it is in the subdirectory
Construction of the directory Book
which is where all the Geometer diagrams used in this chapter are found.
You can learn about the file structure on
the CD by using the Documentation entry in the Help pulldown menu.

C
A

D
Figure 3.2: Bisect a Segment
Construction/Segbisect.T [P]

To step through the construction in Geometer, press the Next button (or type the
n key on the keyboard) to see the next step. The Prev button (or the p key) goes back
one step. (Remember that if the text [P] appears after the file name under a figure, it
means the figure is a proof or construction, and that the Next, Prev, and Start commands
will have an effect.)
The method is shown in figure 3.2. First construct circles centered at A and B,
both having radius AB using CtrEdg=>C. We can do this because we have a point at
the center, and a point on the radius of each of those circles. Those two circles will
meet at points C and D, which we can identify because we can find the points at the

3.5. SIMPLE CLASSICAL EXAMPLES

51

intersections of two circles (command: CC=>P). Connect C and D by a straight line

(command: PP=>L), and that line will intersect the original segment AB at a point M
in the center (command: LL=>P).
In a sense, we are not done; we must prove that the construction in fact bisects the
line segment, so for completeness here is a quick proof (actually, two quick proofs):
Since both circles have the same radius, AC = BC = AD = BD. Therefore
quadrilateral ACBD is a rhombus, and therefore a parallelogram, since a rhombus is
a special type of parallelogram. The diagonals of a parallelogram (AB and CD in this
case) bisect each other so M is the midpoint of AB.
Alternatively, we know that the diagonals of a rhombus are perpendicular so all the
angles at M are right angles. We also know that AC = BC (the circles have equal
radii), and since CM is equal to itself by hypotenuse-leg we know that 4AM C
=
4BM C. Therefore AM = M B so M is the midpoint of the segment AB.

3.5.2 Copy a Segment

Copy a segment AB to an equal-length segQ
A
ment on the line L. Set the compass to the
L
length of AB. Place a mark P on the line L
P
where the new segment is to appear (comB
mand: P on L), and sweep a circle using
P as center and with the compass setting as
above. In Geometer, the command to create
Figure 3.3: Copy a Segment
a circle with a center and a radius defined
Construction/Copyseg.T [P]
by the distance between two other points is
Ctr PP=>C. If Q is one of the points where this new circle crosses L (command:
LC=>P) then the segment P Q has the same length as AB. See figure 3.3

3.5.3 Copy an Angle

Copy an angle ABC so that the vertex of the new angle lies on a line L. See figure 3.4. First pick a point P on L (P on L). Draw a circle centered at B on the original
angle having length AB (command: CtrEdg) and find the point N where it intersects
the other edge of the angle (command: LC=>P). Sweep another circle with radius AB
centered at P on L (command: Ctr PP=>C), and let that circle intersect L at O. Now
construct a circle centered at O of length AN , and the point R where this circle intersects the larger circle centered at P (command: LC=>P) forms an exact copy of the arc
AN . Connect R to P (command: PP=>L) and we are done.
To show that the new angle at P is the same as the old angle at B, note that the two
triangles 4ABN and 4OP R have all three sides equal, and are hence congruent by
SSS. Thus the corresponding angles (ABN and OP R) are equal.

3.5.4 Bisect an Angle

CHAPTER 3. GEOMETRIC CONSTRUCTION

52

Bisect an angle. See figure 3.5. To bisect

ABC, pick a point P on AB, draw a circle with center B through P , and find Q,
B
A
where it intersects BC. Now, from points
P and Q, draw circles with lengths both
R
equal to BP which will intersect at W .
P
O
L
The line BW bisects ABC.
To show that this line does in fact bisect the angle, note that the two triangles
4BQW and 4BP W are congruent, since
BP
= BQ, P W = QW , and BW
Figure 3.4: Copy an Angle
is
equal
to itself. Thus all three sides
Construction/Copyangle.T [P]
are equal to the corresponding sides, so
4BQW
= 4BP W by SSS. The corresponding angles are equal: P BW =
QBW , so the angle P BQ is bisected by BW .

3.5.5 Perpendicular at a Point

Construct a line perpendicular to a given
C
line through a point on that line. See figQ
ure 3.6. To construct a line perpendicular
to L passing through P , first select another
B
W
point Q on L. Draw a circle centered at P
P
passing through Q and let R be the other
A
intersection of this circle with L. Next,
sweep out two larger arcs, say of length
RQ, centered at points R and Q. These
Figure 3.5: Bisect an Angle
arcs intersect at M , and the line P M is the
Construction/AngleBisect.T [P]
required line perpendicular to L and passing through P .
In the figure, 4M P R
= 4M P Q by SSS.
This is because M P is equal to itself, P Q =
R
P R since they are on the same circle centered
P
L Q
at P , and RM = QM since both were drawn
with an equal compass setting (in this case,
equal to RQ). Since 4M P R
= 4M P Q,
M
M P R = M P Q. Since together those an

gles make 180 , they must each be 90 , or right

angles. Thus the line P M is perpendicular to L.
Figure 3.6: Perpendicular to a Line
Construction/PerpPtOnLine.T [P]

3.5.6 Drop Perpendicular to Line

Construct a line perpendicular to a given line
through a point that is not on the line. There is no figure for this one, because in

3.5. SIMPLE CLASSICAL EXAMPLES

53

Construction 3.5.5, figure 3.6 works perfectly well if the point P is not on L. The
proof depends upon the fact that P is on the line, but can be modified to be correct. If
you have doubts, do the construction yourself in Geometer following the written directions for the previous construction and see how it works. Be sure to move P around,
both near and far from the line.

3.5.7 Parallel through a Point

Given a line L and a point P not on L, construct a line through P that is parallel to L.
A figure is not really requireddo the previous construction in Section 3.5.6 twice.
The first time, construct a new line L0 perpendicular to L through P , and then construct
another line through P , but perpendicular to L0 . This second line will be perpendicular
to a perpendicular to L, and will hence be parallel to L, and it was constructed to pass
through P .

3.5.8 Sum and Difference of Angles

To get the sum or difference of angles is easy, assuming you know how to copy an
angle (Construction 3.5.3). Lay out one of the angles, and then put the other one next
to it, either inside or outside, to make the sum or difference.

3.5.9 Triangle Given Three Sides

Construct a triangle given the lengths of the
three sides. This obviously cannot be done
if one segment is longer than or equal to the
sum of the other two, but assuming that is
not the case, the construction is not too hard.
See figure 3.7.

A
A
B

C
C

Copy one of the lengths (say AB) to a

line having new endpoints A0 and B 0 . Then
A
B
draw a circle of radius AC about A0 and an0
other circle of radius BC about B . Call the
intersection C 0 . Triangle 4A0 B 0 C 0 is the Figure 3.7: Triangle from the Lengths
required construction. Because each of the of its Sides
Construction/Triangle.T [P]
sides of the new triangle is an exact copy of
the given lengths, it is clear that 4A0 B 0 C 0
satisfies the conditions.
If you are using a computer geometry program, be sure to change the three given
lengths and see how the construction is modified. Also change them so that they do not
specify a valid triangle (for example, make AB and BC small, and AC large) and see
what happens.

54

CHAPTER 3. GEOMETRIC CONSTRUCTION

3.5.10 Perpendicular Bisector of Segment

Construct the perpendicular bisector of a segment. This is exactly the same as as the
construction in Section 3.5.1, except the result we want is the line, not the point where
the lines intersect.

3.5.11 Circle through Three Points

Given three points, A, B, and C, construct the
B
A
circle that passes through all of them. (See figure 3.8.)
Using the construction from Section 3.5.10
O
C
twice, draw the perpendicular bisectors of both
segment AB and segment BC. Those bisectors will meet at a point O which can then be
used as the center of the required circle.
The reason this works is that every point Figure 3.8: Circle Passing through
on the perpendicular bisector of a segment is Three Points
equidistant from the two endpoints of the seg- Construction/CircFrom3Pts.T [P]
ment. So the point O is the same distance from
A and B, since it is on the perpendicular bisector of AB, and similarly, it is the same distance from B and C. Thus AO = BO and
BO = CO, so AO = BO = CO, so O must be the center of a circle passing through
A, B, and C.

3.5.12 Find a Circles Center

Given a circle, find its center. We basically did
this (and a little more) in Construction 3.5.11.
Simply pick three different points on the given
circle and do the construction above. Any three
different points will work, but if you are doing
this with a physical straightedge and compass, it
is a good idea to pick them far apart on the circle.

C
Figure 3.9: Circle Inscribed in a

3.5.13 Inscribed and Circumscribed Triangle

Construction/Incenter.T [P]
Circles

Given a triangle 4ABC, construct both the inscribed and circumscribed circles.
The circumcircle is the circle passing through
the three points that are the triangles vertices. See figure 3.9 for the construction of
the inscribed circle (called the incircle).

3.5. SIMPLE CLASSICAL EXAMPLES

55

To construct the incircle, construct the angle bisectors at angles A and B. These
will meet at a point O inside the triangle. From O, drop a perpendicular to line AB,
intersecting line AB at point N . The required circle has center at O and passes through
N.
Every point on an angle bisector is equidistant from the two sides of the angle, so
the point O will be equidistant from lines AB, BC, and CA. To find that distance,
we drop a perpendicular from O through any of the sides and find where it intersects
that side. This will be at the shortest distance from O to the side. Since O is this same
distance from all three sides, it is the center of the required circle, and the distance from
O to a side is the radius of the required circle.

3.5.14 Construct Specific Angles

Z
C

Figure 3.10: Construct a 72 Angle

Construction/Seventytwo.T [P]

Construct angles of 90 , 45 , 30 , 60 , and if

you want a challenge, 72 .
The first four constructions are quite simple.
Any time you have constructed a perpendicular
to a line you have constructed a 90 angle. If
you construct a right angle and bisect it, you will
have a 45 angle. To make a 60 angle, just
construct an equilateral triangle (a triangle with
all three sides equal, which you know how to do
since you can construct a circle given the lengths
of all three sides), and all three of its angles will
be 60 . If you then bisect your 60 angle, you

will have one that is 30 .

The construction of the 72 angle is
B
shown in figure 3.10. Begin by laying off a

1
length AB on some line. Construct a perpenT
dicular to the line at B, and a compass cen1
tered at B and passing through A will meet
C
1
1
the new perpendicular at C. Find M , the midpoint of AB, and draw a circle centered at M

that passes through C. This circle will interA

sect the line AB at T . Using the length AT ,
make two circles centered at A and B which
Figure 3.11: The Golden Ratio
intersect at a point Z. The isosceles triangle

Construction/Golden.D [D]
4AZB has two angles measuring 72 and

one measuring 36 .
To understand the construction, let us first look at the properties a triangle with
angles of 72 , 72 , and 36 must satisfy. See figure 3.11. In that figure, angles marked
with the double arcs are 72 and those with a single arc, 36 . If the line AT bisects the
72 angle at A, it will form two 36 angles, as shown.
If we let the side AB measure 1 unit, then AT and CT will also measure 1 unit
because they are parts of isosceles triangles 4ABT and 4CAT . The unknown length

56

CHAPTER 3. GEOMETRIC CONSTRUCTION

is AC, so let us call that . Then the length of BT is 1. By AA, 4ABT 4CAB,
so we have: /1 =
1/( 1). Multiply this out to get 2 1 = 0, whose solution
with > 1 is = ( 5 + 1)/2.
What we have done is to construct , the golden ratio. An isosceles triangle with

two sides equal to and the third equal

to 1 will have two 72 angles in it. The value
of is given by the formula = (1 + 5)/2. To see that the length AT is (assuming
that AB has length 1), note that M B has length
1/2, so by the pythagorean theorem,
the length of M C (which is equal to M T ) is 5/2. AT is the same as M T with 1/2
added on, so AT = .
Note: Since we know how to add and subtract angles and to bisect them, there are
actually quite a few interesting angles that can be constructed from the ones above. For
example, if we subtract a 30 angle from a 36 angle, we can get a 6 angle which can
be bisected to give us 3 . We can then add together any number of 3 angles to get 3 ,
6 , 9 , 12 , . . ..
These are the only angles that can be constructed that are an integral number of
degrees. We know that because if it were possible to get any other angle than a multiple
of 3 then we could subtract the nearest multiple of 3 to get either 2 or 1 . We could
then add together either 10 or 20 of those to construct a 20 angle which we will show
to be impossible in Section 3.12.

3.5.15 Construct a Regular Pentagon

Construct a regular pentagon. Assuming that
C
you can do the construction above to make

4ABC with two 72 and one 36 angle, you

N
M
are almost done, since the exterior angles of
a regular pentagon are all equal to 72 . See
B
A
figure 3.12.
Begin with the triangle constructed in the
previous section and sweep out circles cen- Figure 3.12: Construct a Regular Pentered at points A, B, and C, all having a ra- tagon
Construction/Pentagon.T [P]
dius equal to AB. Let M and N be the points
where these circles intersect outside 4ABC,
and connect them to vertices A, B, and C, as appropriate.
Since all the sides of this figure are equal in length to AB, all we need to do is
show that all the angles are equal as well. In a regular pentagon, all the internal angles
measure 108, and we know that a triangle whose sides are 1, 1, and has angles of
36 , 36 , and 108 , so it is easy to add the angles together in the figure and check that
all of the internal angles of the polygon ABN CM in figure 3.12 are equal to 108 .

3.5.16 Tangent to Circle through Point

Given a circle C and point P on that circle, construct a line through P that is tangent
to C.

3.5. SIMPLE CLASSICAL EXAMPLES

P
O

Figure 3.13: Tangent to Circle at a

Point on the Circle
Construction/Circtan.T [P]

57

This one is not hard. Find the center O of

circle C as we did in the construction described
in Section 3.5.12. Next, construct the line OP .
The line perpendicular to OP passing through
point P is the required tangent line because the
radius of the circle is perpendicular to the tangent line where it intersects the circle.
Obviously if you do the complete construction described above, the figure would be far
more cluttered than what is illustrated in figure 3.13.

3.5.17 Tangent to Circle II

Given a circle C and a point P not on C, construct a line through P tangent to C. See
figure 3.14.
First find O, the center of the circle. Then find M , the midpoint between O and P .
Construct a circle centered at M and passing through O (and P too, of course). The
intersections T1 and T2 of the two circles are the points of tangency, so the required
lines are P T1 and P T2 .
This works because the circle centered at M will have OP as its diameter. For any
point T on that circle, P T O must be a right angle. At the particular points T 1 and
T2 , the right angle will coincide with the boundary of the original circle C, so OT 1 and
OT2 are radii, and hence P T1 and P T2 are tangent lines.

3.5.18 The Geometric Mean

Given two segments with lengths l and m,
construct a segment whose length k is the geometric mean of l and m. In other words, so
that k 2 = l m.

T1
P

One way to find the geometric mean of

T2
two segments is based on the observation that
the altitude to the hypotenuse of a right triangle is the geometric mean of the lengths
into which the hypotenuse is divided. In fig- Figure 3.14: Tangent to a Circle from
ure 3.15, CP is the geometric mean of AP an Exterior Point
Construction/CircPtTan.T [P]
and BP . All three of the triangles in the
figure are similar (4ACP 4CBP
4ABC), so AP/CP = CP/P B which means AP BP = CP 2 .
To construct the geometric mean, draw an arbitrary line and choose a point P on it.
Next, pick points A and B such that P A = l and P B = m. Find the midpoint of the
segment AB and use it as the center of a circle passing through A and B. Thus AB is

CHAPTER 3. GEOMETRIC CONSTRUCTION

58

the diameter, and any point C on the boundary will form a right angle ACB. Draw
a perpendicular to AB through P that intersects the circle at point C If k is the length
of P C, then:
k 2 = l m.

3.5.19 Divide a Segment into n Equal Parts

C

The problem calls for a subdivision of a line

segment into n equal pieces, but the following subdivision into 5 equal pieces can be
modified easily to work for any n. In figure 3.16 n = 5 and we will divide the line
segment AB into 5 equal segments.

First pick a point C not on the line, and

construct line AC. Now pick a point X1
near A, and draw a circle around it passing through A. That circle will also hit in
Construction/Geomean.T [M]
X2 , and clearly, AX1 = X1 X2 . Continue
in this manner, making a circle about X2
which passes through X1 and its intersection with line AC gives a new point X3 .
We can do this as many times as we want to get an evenly spaced sequence of points
X1 , X2 , . . . , Xn along the line. In our case we can stop at X5 since we are subdividing
into 5 parts.

Figure 3.15: The Geometric Mean of

Two Segments

Draw the segment X5 B, and construct

lines parallel to X5 B passing through X4 ,
X3 , X2 , and X1 . These lines will intersect
the segment AB at evenly spaced points
P4 , P3 , P2 , and P1 , which achieves the desired subdivision. The distances between
the Pi are equal because the parallel lines
are evenly spaced.

X3

X4

X5 C

X2
X1
P1 P2 P3 P4 B

Figure 3.16: Dividing a segment into 5

equal pieces

3.5.20 Divide a Line in a Given Ratio Construction/Subdivide.T [P]

If you understood Construction 3.5.19 that divides a line into equal parts, it should be
obvious how to divide it into a ratio. Use exactly the same construction as above, but
when you mark off segments on AC, just mark off two of them, and make them the
relative lengths of the ratios. To prove that your result is correct, look at the two similar
triangles formed.
We will use this trick in the construction of the common internal tangents to two
circles in Section 3.6.1.

3.6. INTERMEDIATE EXAMPLES

59

3.6 Intermediate Examples

3.6.1  Common Tangents to Two Circles
T1
N1
O1 C1

O2 M2 C2
N2
T2

Figure 3.17: External tangents of two circles

Construction/ExtTangent.T [P]

Given two circles C1 and C2 , find the external and internal tangent lines to both of
them. These are really two different problems, so let us begin by finding the common
external tangents. (See figure 3.17.)
Assume that circle C1 is smaller than C2 , otherwise, reverse the names. Find the
centers O1 and O2 of the circles, and then subtract the radii such that M2 C2 = O1 C1 .
Next, construct a circle concentric with C2 , and passing through M2 . Find the tangents to this smaller circle passing through O1 , the center of C1 using the technique of
construction 3.5.17. The common external tangents will be parallel to these tangents,
but pushed out by the radius of C1 . To do this, construct the lines through O2 and the
points of tangency with the circle of radius O2 M2 , and add the radius of C1 beyond the
intersection, giving points T1 and T2 . The parallels to the tangents O1 N1 and O1 N2
passing through T1 and T2 are the required common external tangents.
To see why this is true, notice that the common external tangents are perpendicular
to the radii of both circles. Thus, the distances along the radii of T1 N1 and T2 N2 are
both equal to the radius of C1 .
We will use a different trick to find the common internal tangents. See figure 3.18.
The idea is this: if we look at the solutions in the figure, we see that both the
common internal tangents pass through a point X somewhere on the line between the
centers of the circles. In fact, if we look at 4O1 T1 X and 4O2 T2 X, they are both
right triangles, and O2 XT2 = O1 XT1 since they are vertical angles. That means
4O1 T1 X 4O2 T2 X when X is at the correct point. The similar triangles will have
a ratio of lengths equal to the ratio of the diameters of the two circles, so if we can
divide the line O1 O2 in that ratio at X, that X should be the point we are looking for.
Here is how to do that subdivision, and the proof that the final construction is correct:
Find the centers of the circles and connect them with a line. Then pick an arbitrary
point E not on the line O1 O2 and connect it to O1 . O1 E intersects C1 at P . By
drawing a circle centered at P of radius O2 C2 , we can find a point Q on O1 E such
that P Q = O2 C2 . Connect Q with O2 and construct a line parallel to that line passing
through P . The parallel line through P strikes O1 O2 at X. Finally, using the technique

CHAPTER 3. GEOMETRIC CONSTRUCTION

60

of Construction 3.5.17, construct tangents to C1 passing through X. These lines will

be the common tangents of C1 and C2 .
The reason this works is that for
T1
the
required
internal tangent, trianE
gles 4O1 T1 X and 4O2 T2 X are
P Q
similar, and their sizes are in the
C1
O2 C2
X
O1
ratio of the radii of C1 and C2 .
T2
That is because O1 T1 and O2 T2 are
radii of the circles. Therefore the
point X divides O1 O2 in that ratio.
By picking points P and Q as we
did, the segments O1 P and P Q are
Figure 3.18: Internal Tangents of Two Circles
in the ratio, and the parallel lines
Construction/IntTangent.T [P]
through P and Q must cut O1 O2 in
exactly the same ratio. Therefore a line through X that is tangent to one circle will be
tangent to the other.

T
P

O1
U

X1

O2
X2

Figure 3.19: External tangents by projection

Construction/ExtTangent1.T [P]

Figure 3.19 illustrates that we can also adapt this last method to find the common
external tangents of two circles based on the idea of a projection.
Assume the two circles are centered at points O1 and O2 , as in the figure. A point
X1 is chosen on the circle centered at O1 at any point except on the O1 O2 axis. A line
parallel to O1 X1 is drawn through O2 and it intersects the other circle at a point X2 in
the same direction from O2 as X1 is from O1 .
If the lines O1 O2 and X1 X2 are drawn, they will intersect at a point P that effectively projects the corresponding points of one circle onto the other, including the
points of tangency. The tangents to either circle through P will be tangent to the other
circle. In the figure, the tangents through P to the circle centered at O 1 at T and U are
determined, and those are the required lines.
It is not difficult to show that this works. Begin by showing that 4P O1 X1
P O2 X2 . Next construct a triangle with the same similarity ratio to 4P O1 T and so on.

3.6.2 Equilateral Locus of Points

Let A, B an C be three distinct points in the plane. Find the locus of the centers of
equilateral triangles that can be drawn having one edge passing through A, another

3.6. INTERMEDIATE EXAMPLES

61

through B and the third through C.

When problems like this occur in Euclidean geometry texts, it is likely that the
solution will be something simple: a line or a circle, or perhaps some conic section
like an ellipse or parabola. But even if you know that it is a circle, which circle is it?
The easiest way to proceed is to draw a Geometer diagram and see which points are
generated.
We will not solve this problem here, but will demonstrate how to do the construction, and how to modify the construction to gain a lot more clues about how to approach
the problem.
To construct the Geometer diagram, the first step is easy: create three free points
A, B and C on the plane. Since one of the edges of any equilateral triangle of interest
will pass through the point A, add another free point X and draw the line AX. Since
we dont know exactly where A will sit on that line in the final diagram, change the
line type from segment to line. To construct the locus, we will simply drag the
point X around A and thus sweep out all possible lines.
To construct the other two lines of
the equilateral triangle is not difficult:
we know that they will make 60 anB
gles with the line AX and there are
two
possible ways to do this. The
X
C
A
easiest way is to construct an equilatP
eral triangle using AX as one of the
Q
segments
and the other two sides will
R
make
the
two possible lines at 60 to
P
AX. Its easy to draw an equilateral
triangle; draw the circle centered at A
R
passing through X and then draw the
circle centered at X passing through A.
Figure 3.20: Equilateral Locus
Either intersection of these circles is a
Construction/EquLocus.T [M]
suitable third vertex Y that will form
an equilateral triangle 4AXY . (Note
that either choice of circle-circle intersections will generate a pair of additional lines
that indicate the directions that any line at 60 will to AX.)

Now we simply need to construct lines through B and C that are parallel to the
edges AY and XY of 4AXY , but there are two ways to do this since we can either
draw the parallel to AY through B or C and then the parallel to XY through the other.
If we try both pairings, we find that each of them generates an equilateral triangle, so
there will be two centers associated with each possible line AX.
Label the intersections of the lines to form two triangles: 4P QR and 4P 0 Q0 R0
and all that needs to be done is to find the centers of those triangles. The best way
to locate the centers is probably to find the intersections of a pair of medians for each
or of a pair of altitudes for each.2 A pair of angle bisectors may give bogus results,
since every pair of crossing lines has two angle bisectors and we need to assure that
2 With

Geometer it may be easiest simply to construct the inscribed circle and locate its center.

CHAPTER 3. GEOMETRIC CONSTRUCTION

62

the one chosen is the one running through the triangle. Even if you choose correctly
while drawing the figure, motion of the point X may change this choice, and the other
bisector may result, yielding a bogus triangle center.
At this point, its probably a good idea to change the colors of all the construction
lines to the invisible color, and if youre going to present the drawing to others, you
may want to erase the infinte lines connecting the vertices of triangles 4P QR and
4P 0 Q0 R0 (by making them the invisible color) and redraw them as the segments P Q,
QR, et cetera.
Finally, set the colors of the centers of the triangles to be the smearing color and
drag point X around point A to see what results. The answer appears to be two concentric circles. It is often a good idea to change the point type to dot for locus problems
like this since it is easier to see exactly what shape is swept out by a dot than by one of
the other symbols that indicate points. See figure 3.20.
Notice that in the figure, the points are uniformly drawn around the circles. This is
because the figure was generated from a script that is a bit more complex than the construction described above. The files Construction/EquLocus.T (the original version)
and Construction/EquLocus1.T (the modified version) illustrate how this is done.
Now that we firmly believe that the locus is a pair of concentric circles, the same
figure can be used to find clues about which circles they are. One interesting thing to
do is to change the colors of points P , Q and R (or of P 0 , Q0 and R0 ) to be the smearing
color and see what shapes they sweep out. Try this. Now four circles seem to intersect
at a point. What point is it? See Section 7.4 for more clues.

3.6.3 Three Tangent Circles

A1 A2
B1 B2
C1
C2

Given three different lengths, construct

three mutually tangent circles having
those lengths as their radii. See figure 3.21.

Once you see the trick, this problem

is not too difficult. Notice that if
B
A
the radii are a, b and c, then the distances between the centers of the circles will be a + b, b + c and c + a. Thus
to do the construction, all we need to
Figure 3.21: Three Tangent Circles
do is to construct line segments having
Construction/TangentCircs.T [M]
those three lengths and then construct
a triangle having those segments as the
lengths of its sides. The vertices of the triangles will be the centers of the required
circles.
This can be done on a single line. Pick a point on that line and draw circles of
radius a and b about it. On opposite sides of the point pick the intersections of those
two circles. These points will be a distance of a + b apart. The same line can be used to
compute the distance a + c from what will be the center A of the circle of radius a and

3.6. INTERMEDIATE EXAMPLES

63

the distance b + c from the center of what will be the center B of the circle of radius b.
The intersection of the circle with center A and radius a + c and the circle with center
B and radius b + c will be the center C of the circle with radius c.

3.6.4 Circle Tangent to Four Circles

H
A

G
C

Figure 3.22: Circle Tangent to Four Circles

Construction/Fourcircs.T [M]

Suppose you would like to construct a diagram like that displayed on the left in
figure 3.22. The four inscribed circles are all of equal size.
To perform the construction (see the diagram on the right of figure 3.22), we note
that the centers of the four circles lie on the corners of a square. Those circles will have
radii equal to half the length of the side of the square, and the center of the surrounding
circle will be at the center of the square. With this information, the construction is
fairly straight-forward.

Figure 3.23: Circle Tangent to Five

Circles

First we make a square by drawing a segment AB that will be one side of the square.
Then construct perpendicular lines to the segment AB at both A and B. We then draw circles
about A and B with radii AB, and these circles
will intersect the perpendicular lines at points C
and D that will be the other two corners of the
desired square.

Let E and F be the midpoints of AB and

CD and then the four circles centered at A, B,
C and D passing through E or F will be the
required central mutually-tangent circles. The center of the square is the midpoint of
BC (or of AD or EF ). If we construct a line through B and C it will intersect the
small circle centered at B at Ha point on the desired surrounding circle. Draw that
circle and we are done.
Construction/FiveCircs.T [M]

To create the final figure that appears on the left side of figure 3.22 we simply set
the colors of all the extra construction lines to the invisible color.

CHAPTER 3. GEOMETRIC CONSTRUCTION

64

How would you draw the diagram in

figure 3.23? Hint: remember the construction of the regular pentagon in Section 3.5.15.

3.6.5 Tangent Circle Example

A
X

B
Y

Suppose you would like to draw a diagram Figure 3.24: Circle Passing through Two
where you need to construct a circle that Points and Tangent to a Line
Construction/PPL.T [M]
is simultaneously tangent a line and passes
through two given points. In figure 3.24
we are given the positions of points A and
B as well as the line XY and we would
like to construct the two circles illustrated.
Note: This is a special case of Apollonius problem that requires the construcO1
tion of a circle passing through points or
tangent to lines and circles. The problem
O3
can be stated and solved for any combination of three points, lines and circles,
O2
although there may be multiple solutions.
For the problem of finding all circles that
are mutually tangent to three given circles,
for example, there may be up to eight different
solutions. (See figure 3.25.) If we
Figure 3.25: Eight Circles Mutually Tanlist
the
three given objects with a P for
gent to Three Given Circles
point,
L for line and C for circle,
Construction/Apollonius.T [M]
there are ten different versions of Apollonius problem: (P P P ), (P P L), (P P C), (P LL), (P LC), (P CC), (LLL), (LLC),
(LCC) and (CCC). This particular problem is (P P L). The problem (P P P ) is equivalent to finding the circumcenter of a triangle, and the problem (LLL) is equivalent to
finding a triangles incenter.
This problem probably seems a bit difficult at first, especially since we do not know
how to draw a Geometer diagram to experiment with it. (The reader, of course, can
manipulate the figure to try to learn something about it.)
But even without a Geometer diagram that does the right thing with A, B, X and
Y free, you can at least make an exact drawing by making two circles that intersect and
then drawing a common external tangent to them. If you do this and label the tangent
points as I and J and let P be the intersection of AB with XY , when you test 3 the
fake Geometer diagram in figure 3.26, you will find the following:
Ratios:
[I P] x [J P] = [A P] x [B P]
[J P] x [J P] = [A P] x [B P]
3 See

Chapter 2 to learn how to test a diagram.

3.6. INTERMEDIATE EXAMPLES

65

These relationships are obvious, since

one of the lines from P cuts a secant and
A
the other is a tangent, so P I 2 = P J 2 =
P A P B. It is easy to find P , so P A and
B
P B are available. If we construct P I to
X
I
P
be the geometric mean of those segments
J
Y
as we did in Section 3.5.18 we can find
I. The center of the circle will lie on a
perpendicular to XY through I and on the Figure 3.26: Circle Passing through Two
perpendicular bisector of the segment AB. Points and Tangent to a Line
Construction/PPL1.T [M]
Figure 3.27 shows the complete construction beginning only with A and B and
the line XY . Note: You will almost certainly find the construction easier to follow if
you load the Geometer diagram Construction/LPP2.T and follow along by pressing
the n key.
Draw line AB that intersects XY at P . Draw a circle with center P through B
and find the other intersection of that circle B 0 with line AB. Now P B 0 = P B and
P A is on the other side. Let M be the midpoint of B 0 A and draw a circle through
A with center M . This is the circle with AB 0 as diameter, so the intersection of the
perpendicular to AB through P will intersect that circle at D and the length of P D
will be the geometric mean of P A and P B = P B 0 .
A
O2
O1
B
M
X
I
P
J
D
Y

B
Figure 3.27: Circle through Two Points Tangent to a Line
Construction/PPL2.T [P]

A circle centered at P passing through D intersects the line XY at I and J. The

centers of the required tangent circles will be on the intersections of lines perpendicular
to XY passing though I and J with the perpendicular bisector of the segment AB.
From these centers, O1 and O2 , draw circles passing though A. This completes the
construction.

3.6.6 Another Apollonius Problem

Once we have solved the problem in Section 3.6.5, we can easily solve the (P LL)
version of Apollonius problem: Given a point and two lines, construct the circle that
passes through the point and is at the same time tangent to the two lines. There are
again two solutions for this problem.

66

CHAPTER 3. GEOMETRIC CONSTRUCTION

There are two cases to consider. The

B
easy one is when the two lines are parallel.
P
In that case, the center of the circle must
M
C
C2
1
lie on a line that is parallel to the given
lines and halfway between them. The radius of that circle will be half the distance
A
between the parallel lines, so if we draw a
circle about the point with that radius and Figure 3.28: Circle Passing through a
find the intersections with the middle line, Point and Tangent to Parallel Lines
those intersections represent possible cenConstruction/PLL.T [M]
ters of the required circle. If the point is
between the parallel lines there will be two solutions; if it is outside the parallel lines
there will be no solutions, and if it happens to lie exactly on the line, there will be a
single solution.
The construction for the case of parallel lines is illustrated in figure 3.28. A
P
point A is selected on one of the lines and
B
a perpendicular is drawn through A interP
secting the other line at B. The midpoint
M of segment AB is determined and a line
is drawn through M parallel to the given
lines. Next, a circle of radius AM is drawn
Figure 3.29: Circle Passing through a about the given point P , and the intersecPoint and Tangent to Parallel Lines
tions C1 and C2 of that circle with the line
Construction/PLL1.T [M]
through M are the centers of the required
circles. The required circles have radii of
length AM .
If the lines are not parallel, the solution is actually easier to describe since we
already learned how to solve the P P L problem in Section 3.6.5. This is because if the
lines are not parallel, they will intersect to form an angle. The centers of the required
circles will have to lie on the bisector of that angle, and it is obvious that the required
circles passing through the given point P will also have to pass through the reflection
P 0 of P across the angle bisector. (See figure 3.29.) Now we have the same problem
that was solved in the previous sectionto find a circle passing through two points and
tangent to a line.
Obviously if the point P happens to lie on the angle bisector the problem is also
easy to solve: The line through P perpendicular to the angle bisector will form a triangle and the two required circles are the incircle and one excircle of that triangle.
What happens if P lies on one of the lines?

3.7  Advanced Examples

The solutions to most of the examples in this section are specific to Geometer. If you
use a different application, you will probably need to learn from the users guide how

3.8. PONCELETS THEOREM DEMONSTRATION

67

to solve these problems with that application. It may be useful to have the Geometer
reference manual available as you read this section.

3.7.1 Trisect an Angle

The diagram in figure 3.30 shows what we want: ABC is trisected into three equal
angles by BD and BE. In other words, CBD = DBE = EBA = ABC/3.
If we adjust the size of ABC with the mouse, the sizes of the smaller angles will
change accordingly.
Although this is impossible using classical methods (see Section 3.12), you can trisect an angle with Geometer, but you will have to edit the text version of the diagram.

A
E
D
B

The easiest way to construct the figure

is to use the standard Geometer tools to
draw the points A, B and C, the line segments AB and AC and the angle CBA.
Then edit the diagram, and the first seven
lines of the listing below will be created by
Geometer.

At this point you need to add the next

four lines. The first one that creates af
Construction/GeomTrisect.T [P]
makes a floating point number whose value
is 1/3 of the measure of the angle called
ang1. We then create the angle ang2 having this measure and then add two points (D
and E) making appropriate angles first with segment CB and then with segment CE.
The final two lines can be typed in the editor or created by geometer with the standard
line through two points command.
Figure 3.30: Trisect an Angle

.geometry "version 0.60";

v1 = .free(0.403754, 0.4244, "A");
v2 = .free(-0.796153, -0.531501, "B");
v3 = .free(0.781055, -0.531501, "C");
l1 = .l.vv(v2, v1);
l2 = .l.vv(v2, v3);
ang1 = .a.vvv(v3, v2, v1);
af = .f.rpn(ang1, 3.000000, .div);
ang2 = .a.f(af);
v4 = .v.avv(ang2, v3, v2, "D");
v5 = .v.avv(ang2, v4, v2, "E");
l5 = .l.vv(v2, v4);
l6 = .l.vv(v2, v5);

3.8 Poncelets Theorem Demonstration

68

CHAPTER 3. GEOMETRIC CONSTRUCTION

Poncelets theorem states that if an n-sided

polygon
is at the same time inscribed in one cirB
A
cle and circumscribed about another as is the
quadrilateral in figure 3.31, then if a different
O
starting position is chosen on the outer circle
and lines are successively drawn tangent to the
C
inner circle and extended until they meet the
D
other circle again, this process will terminate
after n steps when the final line returns again
Figure 3.31: Poncelets Theorem for to the starting point. In the Geometer diagram
Quadrilaterals
Construction/Poncelet4.T you can move the
Construction/Poncelet4.T [P]
starting point A around the outer circle and see
that the quadrilateral will close back on A after
four steps no matter where A is located.
In addition, you can move the center point of the inner circle and its size will be adjusted automatically to make an inner circle of the correct diameter that a quadrilateral
will continue to satisfy the conditions of Poncelets theorem.
How is such a Geometer diagram created?
The construction is based on a theorem that states that in the special case of a
quadrilateral inscribed in a circle of radius R and circumscribed about a circle of radius
r, the distance d between the centers of the two circles satisfies:
1
1
1
=
+
.
r2
(R + d)2
(R d)2

(3.1)

With this theorem, the construction of the diagram is straightforward. Here is the
Geometer code that does the trick:
.geometry "version 0.60";
R = .f.rpn(0.950000);
V = .v.ff(0.000000, 0.000000, .in);
C = .c.vf(V, R);
O = .free(-0.125749, -0.269461, "O");
d = .f.vv(O, V);
r = .f.rpn(1.000000, R, d, .add, .dup,
.mul, .div, 1.000000, R, d,
.sub, .dup, .mul, .div, .add,
1.000000, .exch, .div, .sqrt);
c = .c.vf(O, r);
v1 = .vonc(C, -0.884999, 0.345365, "A");
l1 = .l.vc(v1, c, 2);
v2 = .v.lcother(l1, C, v1, "B");
l2 = .l.vc(v2, c, 2);
v3 = .v.lcother(l2, C, v2, "C");
l3 = .l.vc(v3, c, 2);
v4 = .v.lcother(l3, C, v3, "D");
l4 = .l.vv(v4, v1);
l5 = .l.vv(v1, v2);
l6 = .l.vv(v2, v3);
l7 = .l.vv(v3, v4);

3.9. SEVEN TANGENT CIRCLES

69

The outer circle is centered at the origin V with coordinates (0, 0) and its radius is R.
The radius R is a variable since it would be nice to adjust it later to produce a pleasing
diagram where the outer circle is large, but not large enough that the text labels go off
the screen. The value 0.95 was selected by trial and error after the rest of the diagram
was written and debugged.
The point O is the center of the inner circle and can be moved freely by the user.
Once the position of O is determined, the distance d between V and O is calculated.
The next line is a somewhat complex calculation using the reverse-polish calculator
within Geometer to determine the value of r based on the values of R and d using the
equation 3.1. Next, the smaller circle centered at v is drawn with radius r.
At this point, the text editing was terminated and two circles appeared on the Geometer window. Standard Geometer graphical commands were used to construct tangents and intersections of the points B, C and D based on an initial point A constrained
to lie somewhere on the outer circle.
 The Geometer command .v.lcother was used to find the intersections of successive lines with the outer circle since that command guarantees that the other intersection is foundnot the one with the previous point. The standard command .v.lc
could probably have been used, but would not be certain to work in all configurations.
See the Geometer manual for details.

3.9 Seven Tangent Circles

Suppose we would like to make a Geometer diagram like the one shown in figure 3.32. This
seems a lot like the problems in Section 3.6.4
except that it is completely impossible to draw
with classical methods, a regular heptagon.
If you know a little trigonometry, however,
it is not too hard. A point on the unit circle
that makes an angle with the xaxis has coordinates (cos , sin ). We would like to have Figure 3.32: Seven Tangent Circles
Construction/SevenCircs.T [P]
seven evenly spaced points around the origin
(the coordinates of the center of all Geometer diagrams are (0, 0)). Thus we need angles with measures of 0 , (360/7), (2
360/7), . . . The coordinates of the edges of a square Geometer diagram are at x = 1
and y = 1, so if the centers of the seven circles are to be visible, we will need them
to be somewhat less than 1.
Here is the Geometer code that draws the figure:
.geometry "version 0.60";
r = .f.rpn(0.650000);
a0 = .f.rpn(0.000000);
a1 = .f.rpn(360.000000, 1.000000, .mul, 7.000000, .div);
a2 = .f.rpn(360.000000, 2.000000, .mul, 7.000000, .div);

70

CHAPTER 3. GEOMETRIC CONSTRUCTION

a3 = .f.rpn(360.000000, 3.000000, .mul, 7.000000,

a4 = .f.rpn(360.000000, 4.000000, .mul, 7.000000,
a5 = .f.rpn(360.000000, 5.000000, .mul, 7.000000,
a6 = .f.rpn(360.000000, 6.000000, .mul, 7.000000,
x0 = .f.rpn(a0, .cos, r, .mul);
y0 = .f.rpn(a0, .sin, r, .mul);
v0 = .v.ff(x0, y0, .in);
x1 = .f.rpn(a1, .cos, r, .mul);
y1 = .f.rpn(a1, .sin, r, .mul);
v1 = .v.ff(x1, y1, .in);
x2 = .f.rpn(a2, .cos, r, .mul);
y2 = .f.rpn(a2, .sin, r, .mul);
v2 = .v.ff(x2, y2, .in);
x3 = .f.rpn(a3, .cos, r, .mul);
y3 = .f.rpn(a3, .sin, r, .mul);
v3 = .v.ff(x3, y3, .in);
x4 = .f.rpn(a4, .cos, r, .mul);
y4 = .f.rpn(a4, .sin, r, .mul);
v4 = .v.ff(x4, y4, .in);
x5 = .f.rpn(a5, .cos, r, .mul);
y5 = .f.rpn(a5, .sin, r, .mul);
v5 = .v.ff(x5, y5, .in);
x6 = .f.rpn(a6, .cos, r, .mul);
y6 = .f.rpn(a6, .sin, r, .mul);
v6 = .v.ff(x6, y6, .in);
dvv = .f.vv(v0, v1);
rad = .f.rpn(dvv, 2.000000, .div);
c0 = .c.vf(v0, rad);
c1 = .c.vf(v1, rad);
c2 = .c.vf(v2, rad);
c3 = .c.vf(v3, rad);
c4 = .c.vf(v4, rad);
c5 = .c.vf(v5, rad);
c6 = .c.vf(v6, rad);
Rad = .f.rpn(rad, r, .add);
Origin = .v.ff(0.000000, 0.000000, .in);
C = .c.vf(Origin, Rad);
Radin = .f.rpn(rad, r, .sub);
Cin = .c.vf(Origin, Radin);

.div);
.div);
.div);
.div);

The r is the radius of the centers of the seven circles. It is safest to make it a floating
point number called r since if we get it too large or too small, we can just change this
one value later to make it right.
The next seven lines define the angles that divide 360 into seven equal parts. Next
are seven sets of three lines of Geometer code that define the points that are at the
center of each small circle. All of the points are drawn in the invisible color so they
will not appear in the final diagram.
The value dvv is the distance between every pair of centers of those smaller circles,
so the radius of those circles will be rad which is half of dvv.
Seven more lines of code draw the small circles with centers at v0, v1 and so on.
Finally, Rad and Radin are the radii of the outer and inner circles, Origin is the
center of the diagram, and the other two lines of code draw the outer and inner circles.

3.10. A TRICKY CONSTRUCTION

71

3.10 A Tricky Construction

In Section 12.1.11 an interesting problem is presented, and the diagram is based on
the situation illustrated in figure 3.33. Given a circle and an arbitrary chord AB of
the circle, a point X is selected on that chord. The two circles that are simultaneously
tangent to the chord at X and to the original circle are constructed. At the points of
tangency, lines tangent to the circles are drawn. The problem in Section 12.1.11 is to
show that the ratio of the diameters of the two tangent circles is constant, independent
of the location of X, but the problem here is simply to construct the figure. The tangent
lines through E are not necessary for the presentation of the problem, but illustrate
another construction problem that can be tricky.
The easiest way to construct the figure is to use the method of inversion in a circle
covered in Chapter 8. You have probably not read that chapter yet, but reading this
section may convince you that the technique of inversion merits study.
The first part is easy. Draw a circle and then use the P on C command twice to create
points A and B that are constrained to lie on that circle. Draw the segment connecting
A and B and then use the P on L command to create the point X constrained to lie on
AB. So much for the easy part.
We are searching for a line tangent to
a line and a circle and passing through
D
B
a point. If we can somehow invert the
X
problem to send that troublesome point X
A
to infinity then we will be searching for
E
a line instead of a circle in the inverted
problem. To send a point to infinity, we
need to invert in a circle centered at that
C
point.
In this problem, any circle centered at
Figure 3.33: A Tricky Construction
X will do, so pick another point and conConstruction/Tricky.T [M]
struct a circle centered at X and passing
through that point. It may be a good idea
to change the color of that circle temporarily to something you will recognize so that
you do not get mixed up. Since X lies on the line AB, we know that AB will be
inverted into itself, so the only object we need to invert is the original circle.
The command to do that is found in the Primitives pulldown menu under the Circle
submenu. The command is CC=>C Inv (given two circles, invert the first in the second). After issuing the command, click on the original circle (the one with AB as a
chord) and next on the inverting circle. A new circle will appear.
The solutions to this inverted problem will be lines that are parallel to AB and
tangent to the inversion of the original circle. They must be parallel since they will meet
the line AB at infinity (the inverted image of X, the center of the circle of inversion).
There are two such lines.
To find the two solution lines, find the center of the inverted circle, and the easiest way to do that is with the C=>P Ctr command (found in the Primitives pulldown

72

CHAPTER 3. GEOMETRIC CONSTRUCTION

under Point). (Remember that the command creates a point, so it will be in the Point
submenu.) From the point that lies at the center, draw a line perpendicular to AB using
the PL=>L Perp command in the control area. Then find the intersections of this line
with the inverted circle. There will be two solutions. Through both of those points,
construct lines parallel to AB using the PL=>L Par command in the control area and
we will have the two solution lines in the inverted problem.
Finally, to obtain the circles that are mutually tangent to the chord AB at X and
to the original circle, invert each of those lines in the circle of inversion to obtain the
desired tangent circles. Move X, A, and B to test your construction.
At this point, change the color of all the construction lines, points and circles to
the invisible color so that all you can see are the points A, B, X, the segment AB, the
original circle and the two tangent circles. (The names of the points in your diagram
may be different, of course.)
Next, we would like to construct lines that are mutually tangent to the original circle
and to the two tangent circles that you just constructed. The first impulse is to identify
points that are at the intersection of those two circles using the CC=>P command in
the control area, but there are some problems with this. Go ahead and try it and see
what happens as you move points A, B and X.
The problem is that the circles are only constructed to the accuracy of the computer,
not infinitely accurately. Thus the circles that appear tangent may actually pass through
the outer circle twice (very close together), or they may miss it entirely due to numerical
round-off error. As you move A, B and X the intersection points will tend to blink on
and off which would cause the tangent lines also to blink on and off as the diagram is
modified.
The best solution to get a solid diagram is this. The point D in the diagram must
lie on the radius of both circles passing through that point. Thus the centers of both
circles lie on a line passing through D. Find the centers of both circles. Depending on
how you constructed the original circle, its center may be there, but simply colored the
invisible color. If that is the case, click on the Show Invis button to see it, select it, and
change its color temporarily.
When you have both circle centers, connect them with a line and find the intersection of that line with the outer circle. Do the same thing to find the point C. Be careful
to find the intersection with the outer circle. We will see why shortly.
Finally, construct tangent lines to the outer circle using the PC=>L command in
the control area, selecting the point and the outer circle. Find the intersection of the
two lines at E and you are done.
If you were not careful and selected D to be the intersection with the inner circle,
it might be slightly inside the outer circle from the point of view of the computer since
the computers mathematics is not perfect. If it is inside by a millionth of a centimeter,
there will still be no tangent line. You could, of course, have selected C and D to be on
the inner circles, and then construct tangents through those points to the inner circles.
Either way works fine; the only way to have problems is to select the points on one
circle and then draw tangents to the other. It may be worth experimenting with the
incorrect constructions to see the sorts of problems you have: lines that blink on and
off again.

3.11. CLASSICAL CONSTRUCTION

73

3.11 Classical Construction

The classical constructions use only a straightedge and compass and a precise set of
rules about how they may be used. In almost every case, you are given certain lengths,
angles, points, lines, or circles, and are asked to construct others using only a straightedge and compass. See [Martin, 1991] for a book completely devoted to geometric
construction.
Geometer and other computer geometry programs make this much easier than it is
with a physical compass and straightedgethe lines are more precise, any part of the
drawing can be adjusted as the construction progresses, and it is easy to undo errors.
Besides, using Geometers proof technique, any of the sample constructions in this
chapter may be viewed step by step much more easily with Geometer than by looking
at the diagrams in the text.
The problem statement in the first paragraph of this section is sloppy: what does
it mean when we say, using only a straightedge and compass? Are you allowed to
put marks on the straightedge? Can you make a tangent to a circle by placing the
straightedge against it so that it just touches? Even if you are not allowed to mark the
straightedge, could you pin the compass to it with your fingers so that you can at least
effectively have two marks?
Here is a precise definition of the meaning of construction with a straightedge and
compass.
At any stage in the construction, you may do any of the following things to obtain
additional points, lines or circles:
1. You may draw a straight line of any length through two existing points.
2. You may find a new point at the intersection of two lines, of two circles, or of
a line and a circle. When you are given a segment, of course, you are given the
two endpoints, so you can certainly use those.
3. You may construct a circle centered at any existing point having a radius equal
to the distance between any two existing points. In other words, you can set the
size of the compass from any two points A and B, and then you can move the
point of the compass to another point C without changing the setting and draw a
circle of radius AB about the point C.4 (Of course this includes drawing a circle
given its center and a point on the edgeyou use the center and the edge to set
the compass size, and then you re-use the center point as the center of the circle.)
4. You may choose an arbitrary point on a line, circle, or on the plane. (And of
course you can also choose a point not on a line or circle as in pick any point
not on segment AB.)
4 This is a modern compass. Originally, most construction problems were stated in terms of a Euclidean compass which collapses after each circle is drawn. In other words, a Euclidean compass cannot be
used to copy a length. But it is possible to do any construction with a Euclidean compass that can be done
with a modern compass. In fact, all constructions that can be done with a straightedge and compass can be
done with a compass alonethis is called the Mohr-Mascheroni theorem. See [Martin, 1991].

74

CHAPTER 3. GEOMETRIC CONSTRUCTION

In the real world of pencils and paper (or computer calculations) none of these
constructions can be carried out exactly for arbitrary points. But a mathematical construction assumes perfect toolsthat you can find exactly where two lines cross, and
you can set a compass width to exactly the distance between two points.
On the other side of the coin, given these perfect tools, you must find the required
point or length exactly. Getting the answer accurate to one part in a trillion is not good
enough. If all that is required is an answer that is only accurate to, say, one part in a
trillion, there is usually a totally mechanical way to proceed, which will be discussed
later.

3.12  Some Constructions Are Impossible

Many, many constructions using only a straightedge and compass are possible. In fact,
it seemed to the ancient Greeks that almost any length could be constructed, but from
antiquity, there were three famous problems that seemed to defy solution:
Trisection of an angle: Given an arbitrary angle , construct a new angle having onethird the size: /3.
Doubling a cube: Given the length of the edge of a perfect cube, find the length of an
edge of a cube having exactly twice the volume of the original cube.
Squaring the circle: Given a circle, construct a square having exactly the same area as
the circle.
One by one (but thousands of years later) all three of these constructions were
proved to be impossible. But the problem statements are so easy and the proofs of
impossibility so difficult to understand, especially in the case of squaring the circle,
that every day letters arrive at mathematics departments throughout the world from
crackpots claiming to have solved one of these three classic problems (usually angle
trisection)5 .
The constructions that are claimed to solve one of the problems above invariably
contain an error. Quite often, however, the construction is correct in a sensethe
error is that the solver did not understand the valid construction operations stated at
the beginning of this chapter6 . If you can put two marks on your straightedge, it is easy
to trisect any angle, for example. Figure 3.34 shows how.

3.12.1 Angle Trisection

See figure 3.34. Construct an arbitrary circle with the angle that you wish to trisect
as a central angle. Then put two marks on your straightedge the same distance apart
5 See [Dudley, 1994] for a huge collection of incorrect methods for angle trisections submitted by these
crackpots.
6 For example, instead of straightedge and compass one often hears the expression ruler and compass
and the word ruler may imply that some sort of measurement is possible.

3.12. SOME CONSTRUCTIONS ARE IMPOSSIBLE

75

N
M

/3

2/3
/3

Figure 3.34: Trisecting an Angle

Construction/Trisect.T [M]

as the radius of the circle. Position the straightedge so that one of the marks is on the
extension of the line AB (at M ), the other is on the circle (at N , so M N = AB), and
so that the straightedge passes through the point C where the central angle intersects
the circle. Then AM C = /3.
The proof is not difficult. Let denote the unknown angle at AM C, and construct
the line AN . 4ANM is isosceles, so NAM = . Since a straight angle measures
180, we have:
+ NAC + = 180.
(3.2)
Since AN = AC, 4NAC is also isosceles, with AN C = 2 because the exterior angle of a triangle is equal to the sum of the other two interior angles. Therefore,
NAC = 180 4. We can substitute this value for NAC in equation 3.2 above
and get
+ 180 4 + = 180,
from which it is easy to conclude that = 3, or = /3 so we have achieved the
trisection, but it required marking the straightedge.

3.12.2  Impossibility Proofs

The proofs that the three problems are impossible to solve using the standard rules for
straightedge and compass construction follow the same pattern. First, you show that the
given construction operations basically can only generate new points that are related
to the old points via arithmetic operations: (+, , , and ), or the square root. Then
you show that certain numbers cannot be achieved in this waycertain cube roots,
for example, or , since it is transcendental (in other words, it is not the solution to
any non-zero algebraic polynomial equation with integer coefficients, no matter how
complicated). Then doubling the cube and squaring
the circle are clearly impossible

3
(doubling the cube
requires
the
calculation
of
2
and
squaring the circle requires the

construction of ). Angles cannot be trisected in general because the construction

of a 20 angle implies that you can find the cosine of 20 , which is the solution of an
irreducible cubic equation over the integers. (To see this, expand cos 3 in terms of
cos ):

76

CHAPTER 3. GEOMETRIC CONSTRUCTION

cos 3 = cos(2 + ) = cos 2 cos sin 2 sin

= cos( + ) cos sin( + ) sin

= (cos2 sin2 ) cos 2 sin cos sin

= cos3 3 sin2 cos
= cos3 3(1 cos2 ) cos
= 4 cos3 3 cos .

If 3 = 60 , cos 3 = 1/2, so the equation for cos is 1/2 = 4 cos3 3 cos ,

which is a cubic equation whose solution requires taking a cube root. In other words,
if x = cos 20 , then x is a solution to the equation
8x3 6x 1 = 0.
Therefore, one cannot construct the cosine of 20 , and hence one cannot construct
a 20 angle. Since there is one particular angle that cannot be trisected, the problem
has no general solution. (Remember that some angles can be trisectedfor example, it
is pretty easy to trisect a 90 angle. In fact, it is often the case that the crackpot angle
trisection solutions really trisect a single special angle.)

3.12.3  Constructible Regular Polygons

There are some oddball regular polygons that can be constructed. Without proof,
here is the general formula. (See [Martin, 1991] for the details.)
If p1 , p2 , . . . , pn are distinct Fermat primes and k is any non-negative integer, then
a regular polygon with m = 2k p1 p2 pn sides can be constructed with a straightedge
i
and compass. A Fermat prime is a prime number of the form 22 + 1, where i is an
integer. At the current time, only five different Fermat primes are known: 3, 5, 17, 257,
and 65537, corresponding to i values of 0, 1, 2, 3, and 4. Thus it is possible to construct
the regular 17-gon (called a regular heptadecagon), and such a construction (but not a
proof that the construction is correct) is shown in Section 2.1.3. But (in theory, at least),
it is also possible to construct a regular 257-gon and a regular 65537-gon.
If you believe that it is possible to construct a regular equilateral triangle, a regular
pentagon, et cetera, it is not hard to convince yourself that the other constructions of the
form above can also be done. It is not so easy to show that these are the only possible
ones.
For example, to construct a regular 60-gon (60 = 22 3 5), construct a regular
3-gon (an equilateral triangle) and a regular 5-gon (a regular pentagon), and subtract
the central angles. One is a third of a circle and the other is a fifth of a circle, so the
difference is an angle that is 1/3 1/5 = 2/15 of a circle. Bisect that angle three times
to get an angle that is 1/60 of a circle.

3.13. 139 MORE PROBLEMS

77

3.13 139 More Problems

In [Wernick, 1982], William Wernick provided a list of 139 triangle construction problems and gave the solutions for many of them. A more recent update on the state of the
problems appears in [Meyers, 1996].
Basically, all require the construction of a triangle given three located points chosen
from among the following:
A, B, C, O
Ma , M b , M c , G
Ha , H b , H c , H
Ta , T b , T c , I

the three points, and circumcenter

feet of the three medians, and centroid
feet of the three altitudes, and orthocenter
feet of the internal angle bisectors, and incenter

In other words, a problem like (A, Ma , G) means Construct the triangle 4ABC
given the points that identify the point A, the foot of the median beginning at A, and
the location of the centroid.
At the time of publication of [Meyers, 1996], there were still 20 open problems,
among them, the following: (O, Ha , Tb ), (O, Ha , I), (O, Ta , Tb ), (Ma , Mb , I).
Some of the constructions are impossible (for example (G, Ha , Hb ), since it is
equivalent to finding the root of an irreducible cubic equation) and some are trivial (for
example (A, B, C)).

3.14 Construction Exercises

The following exercises do not have solutions printed in the text. Unless they are
marked with a warning  symbol, they are not too difficult. You can find the solutions
in Geometer files having the same name as the exercise. For example, to find the
solution to problem C15, view the file C15.T in the Construction directory.
1. [C1] Given an isosceles right triangle 4ABC where BAC = 90 , find points
H, I, J, and K such that H and I lie on BC, J lies on AB, and K lies on AC
such that the quadrilateral HIJK is a square.
2. [C2] Given a semicircle centered at a point C with diameter AB, find points I
and J on AB, and points H and G on the semicircle such that the quadrilateral
GHIJ is a square.
3. [C3] Given a quadrant of a circle (two radii that make an angle of 90 and the
included arc), construct a new circle that is inscribed in the quadrant (in other
words, the new circle is tangent to both rays and to the quarter arc of the quadrant).
4. [C4] Given a point A, a line L that does not pass through A, and a point B on L,
construct a circle passing through A that is tangent to L at the point B.
5. [C5]  Given a point A, a line L, and a length x, construct a circle passing
through A of radius x that is tangent to the line L.

78

CHAPTER 3. GEOMETRIC CONSTRUCTION

6. [C6]  Given a circle K, a line L, and a point A on L, construct a circle that is

tangent to K and tangent to L at the point A. (Hint: There may be more than
one solution.)
7. [C7] Given a point A and two straight lines L1 and L2 , construct a line L that
passes through A, and so that the angle L makes with L1 is the same angle that
it makes with L2 .
8. [C8] Given two points A and B and a line L, find a point C on L such that
AC = BC.
9. [C9] Given two points A and B that both lie on the same side of a line L, find a
point C on L such that AC and BC make the same angle with L.
10. [C10] Given two points A and B that both lie on the same side of line L, find a
point C on L such that AC + BC is as small as possible. (Hint: This problem is
related to the construction problem 9. Also remember that the shortest distance
between two points is a line.)
11. [C11] Given a point A on a line L. If B is any point on L, divide the segment
AB into a given ratio by another point C (in other words, find C so that AC :
BC = r for some fixed ratio r). What is the locus of all possible positions of the
point C?
12. [C12] Trisect a right angle.
13. [C13]  If A and B lie on a line L, the two points C and D on L are said to
harmonically divide the segment AB if AC/CB = AD/DB. Given points A,
B, and C on a line L, find the point D such that C and D harmonically divide
AB.
14. [C14] Given three points, A, B, and C on the plane, but not collinear, find
another point X such that if L is any line through X, and if perpendiculars are
dropped to L from A, B, and C that meet L at A0 , B 0 , and C 0 , respectively, then
AA0 + BB 0 = CC 0 . (Hint: Can you find any particular lines that work?)
15. [C15] Given two non-parallel lines L1 and L2 and a radius r, construct a circle
of radius r that is tangent to both L1 and L2 .
16. [C16] Given a line L and two points A and B on the same side of L and not on
L, and given an angle , construct a point C on L such that ACB = . For
certain configurations there may be no solutions.
17. [C17] Given a circle K, a point P , and a segment AB whose length is less than
the diameter of K, construct a line L through P which intersects K at P1 and P2
such that AB = P1 P2 .
18. [C18]  Given two circles K1 and K2 that do not intersect, and such that neither
lies inside the other, together with a line segment AB such that the length of AB
is smaller than the diameter of the smaller circle, construct a line passing through
both circles such that both of the chords cut by the circles have length equal to
AB.
19. [C19] Given a circle K and two perpendicular radii of K, construct a line that
cuts the K twice and each chord so that the three segments of the line between
these intersections have equal length.
20. [C20]  Given a triangle and a point on one of the sides of the triangle, construct
a line through that point that divides the area of the triangle in half.

3.14. CONSTRUCTION EXERCISES

79

21. [C21] Given three lines in the plane, find all the circles that are simultaneously
tangent to all three lines.
22. [C22] Given any triangle 4ABC, construct the point P inside 4ABC such that
A(4AP B) = A(4BP C) = A(4CP A).
23. [C23] Given a parallelogram and a point on one of its edges. Construct a line
through that point that bisects the area of the parallelogram.
24. [C24]  Given a convex quadrilateral ABCD, construct a line through A that
bisects the area of the quadrilateral.
25. [C25]  Given a convex quadrilateral ABCD and a point P on segment AB,
construct a line through P that bisects the area of ABCD.
26. [C26] Given a line L and two points P and Q on the same side of L, construct
the point R on L such that the angle P RQ is as large as possible.
27. [C27] Given a circle K and two points P and Q on the outside of K, construct
the point R on K such that the angle P RQ is as large as possible.
28. [C28]  Given an isosceles triangle, construct an equilateral triangle of equal
area.
29. [C29]  Given the two radii R and r, and a straight line L, construct two circles
having those radii that are externally tangent to each other and are externally
tangent to L.
30. [C30] Construct a circle that passes through two given points and is tangent to a
given circle.
31. [C31] Given a point A, a circle K, and a point P on K, construct a circle passing
through A that is tangent to K at P .
32. [C32] Given a circle K, a line L, and an angle , construct a line tangent to K
which makes an angle with L.
33. [C33] Suppose you are given two non-parallel lines that meet so far away that
you cannot access the point of intersection. Construct the bisector of the angle
formed by those two lines.
34. [C34] Given a circle K and two points P and Q. Construct a circle K 1 passing
through P and Q such that K1 intersects K at two diametrically opposite points
of K.
35. [C35]  Given the hypotenuse and area of a right triangle, construct the triangle.
36. [C36] Given two points, construct two equal circles centered at those points such
that their common external tangent passes through a third given point.
37. [C37] Given two points, construct two equal circles centered at those points such
that their common internal tangent passes through a third given point.
38. [C38] Given two points and a circle, construct two circles of equal radius with
centers at the two points such that the common external tangent of the circles is
tangent to the given circle.
39. [C39]  Given two concentric circles and a point P between them, construct the
circles tangent to both concentric circles and to P . What happens if P is outside
the outer circle? Inside the inner circle?
40. [C40]  Given three mutually tangent circles, construct the two Soddy circles
that are tangent to all three of them. If the triangle whose vertices are the centers
of the three circles, the centers of the Soddy circles are called the Soddy circles,
centers.

80

CHAPTER 3. GEOMETRIC CONSTRUCTION

41. [C41] Given a right triangle and one of its legs, construct a point on that leg that
is equidistant from the hypotenuse and the vertex opposite the hypotenuse.
42. [C42] Given three points, construct a parallelogram such that the three points are
the midpoints of three of its sides.
43. [C43] Given a circle, a line, and two points, construct a circle passing through
the two points whose common chord with the given circle is parallel to the given
line.
44. [C44] Construct 4ABC given the length of side BC, the length of the bisector
of ABC, and the size of ABC.
45. [C45]  Given segments of length x, y, and a, construct a segment of length b
such that x : y = a2 : b2 .
46. [C46] Construct 4ABC given BC, AC, and the length of the median from A
to the midpoint of BC.
47. [C47]  Construct a quadrilateral ABCD given A, B, C, AB, and AD.
48. [C48] Given two equal intersecting circles and an angle, construct two equal
chords of the two circles that meet at a point of intersection of the circles and
form an angle equal to the given angle.
49. [C49] Given 4ABC construct its two Brocard points. A Brocard point is a
point that makes equal angles with the sides of the triangle at the vertices. In
other words, find a point such that CA = AB = BC. The other
Brocard point 0 makes the other angles equal: BA0 = AC0 = CB0 .
50. [C50] Construct the midpoint M of segment AB using a compass alone.
51. [C51]  Construct a triangle given the lengths of its three altitudes. (Hint: The
area of a triangle is half the base multiplied by the altitude. But you can use any
altitude.)
52. [C52] Construct a triangle given the locations of the feet of its altitudes. (Hint:
Look at the classical triangle centers of the triangle connecting the feet of the
altitudes.)
53. [C53] Construct a triangle given one of its angles together with the radii of the
incircle and the circumcircle. (Hint: See Section 6.1.)

Chapter

Computer-Aided Proof
At the time this book is published, commercially available computer geometry programs are not capable of constructing the proof of a geometric theorem, but they can
be used to aid in that process. In this chapter we will examine situations where tests
and experiments can be done using Geometer to help construct a proof of geometric
theorems.
Perhaps the most common use of computer geometry programs is to search for
theorems and to provide evidence that a theorem or relationship is true. If you are told
that the three altitudes of a triangle are concurrent at a point and you do not believe it,
you can draw a triangle and its altitudes, and by moving the vertices of the triangle you
can see that (at least to the accuracy of the computer screen and your vision) that they
always seem meet at a point in thousands of cases. This is not a proof, of course, but
when one is convinced that a result is true, the proof is psychologically easier to find 1 .
1A

great example of this is the firing squad synchronization problem which is described in

81

82

CHAPTER 4. COMPUTER-AIDED PROOF

Warning: All of the examples in this chapter rely on Geometer. Other computer
geometry programs may have similar commands, but if you use one of those programs,
you may need to have the Geometer reference manual available to fully understand all
of the examples.

4.1 Finding the Steps in a Proof

At this point it might be good to review the example presented in Section 2.8.
A formal proof in geometry (and in mathematics in general) consists of a series of
statements beginning with the given information, and where each statement after that
is a logical conclusion made from some combination of the previous statements. Every
statement in the proof will be a consequence of the initial information, and the final
statement is the theorem which was to be proved.
Proofs that appear in mathematical texts and journals may differ from this somewhat to make them easier to understand, but it should be clear how to convert them into
formal proofs. For example, quite often you will see something like this: If we can
show that X is true, then our theorem is true because . . . . Later, there is a line like
this: So since we now know that X is true, we are finished. The formal proof would
contain those statements in the opposite order, but the rearrangement above probably
makes it easier for the reader to see where the argument is going.
The idea of a formal proof is driven home by the T proofs required in most high
school geometry classes. Students draw a giant T on their paper, and every statement
on the left has a corresponding reason on the right why it is true. The final statement
on the left is the theorem to be proved.
For any interesting theorem, the proof does not come out of a mathematicians head
that way. A bunch of things seem to be true, some depending on others, and the job
of constructing a proof sometimes simply amounts to sorting these into a logical order.
More often, however, it is difficult to discover many of those true things.
One of the most powerful features of Geometer is that it can often provide a long
list of possible entries for the left column of a T proof. Of course you then need to
sort them into a logical order and provide reasons why each is true.

4.2 Testing a Diagram

Geometer contains a pair of commands, Test Diagram and End Test that provide a list
of relationships that seem to be true.
[Minsky, 1967]. The problem remained an unsolved research problem for years, but as soon as the first
solution was found, it became a standard, not too difficult exercise in computer science texts. Another good
example is the following problem: Prove that the number between any pair of twin primes except the pair
(3, 5) is divisible by 6. Most mathematicians panic at this point, having no idea how to proceed since so
little is known about twin primes. But as soon as you add, You have 10 seconds to solve it, almost all can
prove it instantly.

4.3. ALTITUDES AS BISECTORS

83

To use this tool, after a diagram is drawn in Geometer, the Test Diagram button
is pushed. Then the diagram can be modified at will, and the test is ended with the
End Test command. At the end of the test, Geometer presents a list of all the relationships it discovered that seemed to hold throughout the diagram modifications. These
include such things as equal segments, equal angles, concurrent lines, collinear points,
parallel or perpendicular lines, and various other things.
Geometer makes an effort to avoid listing obvious relationships. In other words,
if point C is constrained to lie on the line connecting points A and B, Geometer will
not report that A, B and C are collinear. They are collinear by definition. Similarly,
equal vertical angles will not be reported, et cetera. Geometer is conservative, however, so some of the things on the list may be obviously true to a human, but they
were not obvious to Geometer.
The command pair that tests diagrams works like this: When Geometer enters the
testing mode, it examines all sets of three points to see which sets all lie on the same
line, it looks at all sets of four points to see which sets all lie on the same circle, and so
on for all the other sorts of relationships it seeks. When any of these occur, and are not
obviously true, Geometer adds them to a list.
After making the initial list, as the diagram is modified while in testing mode, those
relationships are tested over and over. If a relationship fails, it is removed from the list 2 .
This is because a relationship may have held by accident in the initial configuration, but
does it does not hold in general. When the user indicates that the test is over, Geometer
presents everything that remains on the list3 .

4.3 Altitudes as Bisectors

B

C
Hc
A

H Ha A
Hb
C
B

Figure 4.1: Altitudes as Bisectors

Proofaid/AltBisect.D [M]

some others as well:

Let us begin with a very easy example

where we will illustrate a relationship
between the altitudes of one triangle and
the angle bisectors of a related triangle.
Show that if the altitudes of 4ABC
are extended to intersect that triangles
circumcircle at A0 , B 0 and C 0 , forming
4A0 B 0 C 0 as in figure 4.1, then the altitudes of 4ABC are the angle bisectors
of 4A0 B 0 C 0 .
If we test the diagram with Geometer, we obtain a list of relationships, but
the important ones are these that include
not only the angles that interest us, but

2 In practice, usually only a tiny motion of a couple of points is needed to eliminate relationships that
occur by accident
3 With suitable editing, of course. For example, if A, B and C are collinear, and so are A, B, and D,
then Geometer will report that the four points A, B, C, and D are collinear.

CHAPTER 4. COMPUTER-AIDED PROOF

84

Equal angles:
(B A A) (B C C) (B B A) (B B C)
Equal angles:
(C B B) (C A A) (C C A) (C C B)
Equal angles:
(A B B) (A C C) (A A C) (A A B)

If we simply consider the first set of four equal angles, it is obvious that BAA 0 =
BB 0 A0 and that BCC 0 = BB 0 C 0 since both angles in each pair are inscribed in
the same arc of the circle.
There are many ways to proceed, but let us begin with an easy one. We see that
HHa C = HHc A = 90 since AH and CH are altitudes, and Ha HC =
Hc HA because they are vertical angles. Thus 4Ha HC 4Hc HA because of
AA, and so BCC 0 = BAA0 . Thus all four angles in the previous paragraph are
equal, so B 0 B bisects A0 B 0 C 0 . Since there is nothing special about the angles we
chose (or by symmetry, if you prefer to say it that way) the same is true of all the other
angle bisectors.

4.4 Another Bisector

Two circles intersect at points M and N .
Q
The common tangent P Q is drawn on
P N
the side of the circles nearest N as in figR
ure 4.2. The line P N is extended until
it intersects the other circle at R. Show
that M Q bisects P M R.
M
This time if we test the diagram,
there is absolutely no useful information returned that might help us with
Figure 4.2: Another Bisector
the proof, other than reporting that M Q
Proofaid/CircBisect.D [M]
does, in fact, seem to bisect P M R.
When this happens a good strategy is to look at the figure to see if drawing some
auxiliary lines might help. This is such a simple figure that there are not too many
possibilities, so we might as well draw the obvious ones: M N , N Q and QR. In more
complex figures it is usually a bad idea just to draw every line you can think of, since
the number of reported relationships can easily become unwieldy.
But with the addition of just those three additional lines, the list of relationships
now includes a few that might be useful:
Equal angles:
(N R M) (M
Equal angles:
(P Q N) (N
Equal angles:
(Q P M) (M
Equal angles:
(Q P N) (P

Q N)
R Q) (Q M N)
Q R)
M N)

4.5. PERPENDICULAR DIAGONALS

85

There are a few ways to complete the proof with this information, but probably
the easiest is simply to use the last equality: QP N = P M N . The angles are
obviously equal since P Q is tangent to the smaller circle in the figure and cuts off
exactly the same arc N P as does N M P .


We know that angle QP N is equal to the difference of arcs RQ QN. We

also know that QN = QM N and that arc RQ= RM Q, so if we combine all this
information we have:
N M P = QP N = RM Q QM N,
so QM P = QP N + QM N = RM Q which means that M Q bisects P M R.

4.5 Perpendicular Diagonals

D

A
J

O
B

L
C

Figure 4.3: Perpendicular Diagonals

Proofaid/PerpDiags.D [M]

If a convex quadrilateral ABCD has perpendicular diagonals that meet at a point

O, if the quadrilateral is rotated about O by an arbitrary angle to form the new quadrilateral A0 B 0 C 0 D0 , then the points J, K, L and M that lie, respectively, at the intersections of AB and A0 B 0 , BC and B 0 C 0 , CD and C 0 D0 and finally DA and D 0 A0 all lie
on a circle. See figure 4.3.
When the diagram above is drawn in Geometer and then is tested for relationships
that hold as the figure is manipulated, Geometer returns a long list. Many items in the
list are obvious, but the following jump out at us.
Points
O
Points
O
Points
O
Points
O

on
D
on
B
on
C
on
A

a circle:
D L M
a circle:
B J K
a circle:
C K L
a circle:
A J M

CHAPTER 4. COMPUTER-AIDED PROOF

86

Now that it is pointed out to us that these sets of five points lie on circles, it is easy
to see why. To take the first one as an example, since the two quadrilaterals are rotated
by an angle relative to each other, DM D 0 , DOD0 and DLD0 are all equal to
and therefore lie on the same circle.
Since they lie on the same circles, there are other angles in those circles subtended
by equal arcs, so we can write eight equations like: ODM = OLM , ODL =
OM L and so on.
Since the diagonals of the quadrilateral are perpendicular, we know that triangles
4BOA, 4AOD, 4DOC and 4COB are right triangles so their non-right angles are
complementary.
Here is a set of equations derived from the observations above:
OJK

= OBC

(4.1)

OJM
OLK

= OAD
= OCB

(4.2)
(4.3)

OLM
ODA + OAD

= ODA
= 90

(4.4)
(4.5)

OBC + OCB

= 90

(4.6)

Substituting equations 4.1, 4.2, 4.3 and 4.4 into equations 4.5 and 4.6, and adding
those equations together yields:
OLM + OJM + OJK + OLK = 180 .

(4.7)

But since OLM + OLK = M LK and OJM + OJK = KJM , equation 4.7 tells us that M LK and KJM are complementary, so JKLM must be a
cyclic quadrilateral.

4.6 Constant Sum

Show that if any point is picked in an equilateral triangle that the sum of the distances to
the three sides is constant.

C
G
F
c Db
A

a
E

Figure 4.4: Equilateral Projections

Proofaid/Equilateral.T [M]

To make a Geometer diagram, everything is easy until the final step. We can
draw an equilateral triangle beginning with
one side by drawing circles centered at each
end of the arc and passing through the opposite end. The intersections of the circles are
suitable positions for the third point, that can
be connected, giving us 4ABC. Then place
a free point D inside and drop perpendiculars

4.6. CONSTANT SUM

87

to each side. Geometer can only calculate the distances between points, so find the intersections, and you should have a diagram like that in figure 4.4. It would probably be
better to give the circles the invisible color to simplify the final diagram.
Now we would like to display the lengths of the three segments DE, DF and
DG. This is also not hard. Select one of the lines and then choose the Display Value
command from the Edit pulldown menu, you will get something like l4 = 0.425
displayed in the upper left of the Geometer window. So far you have not given that
line a name (points receive names by default but other geometric figures do not). You
can do so by selecting it (it should still be selected if you just issued the Display Value
command) and issue the Edit Name command in the Edit pulldown. This brings up a
dialog box into which you can type the name of the line. In figure 4.4, the lines have
been named a, b and c and the display for each has been toggled on. (The keyboard
shortcut for those commands is ctrl-T to toggle the display and ctrl-n to change the
name.)
When the lengths are displayed, they appear in the upper left of the diagram like
this: a = 0.514.
As you move point D around, the values of the lengths are continuously updated,
but it is still a bit of work to add the three each time to assure yourself that the sum is
constant. Why not have Geometer do that for you? To do this you will need to edit the
text version of the file. The entire modified file is listed below since you need to know
the internal names that Geometer gave to the various points before you can type in the
final code.
.geometry "version 0.60";
v1 = .free(-0.92, -0.871156, "A");
v2 = .free(0.893333, -0.871156, "B");
c1 = .c.vv(v1, v2);
c2 = .c.vv(v2, v1);
v3 = .v.cc(c1, c2, 2, "C");
l1 = .l.vv(v1, v2);
l2 = .l.vv(v2, v3);
l3 = .l.vv(v3, v1);
v4 = .free(0.026616, -0.372624, "D");
l4 = .l.vlperp(v4, l1, .display, "a");
l5 = .l.vlperp(v4, l2, .display, "b");
l6 = .l.vlperp(v4, l3, .display, "c");
v5 = .v.ll(l1, l4, "E");
v6 = .v.ll(l2, l5, "F");
v7 = .v.ll(l3, l6, "G");
DE = .f.vv(v4, v5);
DF = .f.vv(v4, v6);
DG = .f.vv(v4, v7);
S = .f.rpn(DE, DF, DG, .add, .add, "Sum");

The three lines beginning with DE, DF and DG set those names to be the distance
between the points in parentheses. The function .f.vv makes a floating point number
(.f) from the distance between two points (.vv). The final line performs a floating point
calculation with Geometers built-in rpn4 calculator. The three lengths are pushed on
4 Reverse

Polish Notation.

CHAPTER 4. COMPUTER-AIDED PROOF

88

the stack, and then the top two items are added and the operation repeated. Since the
floating point number S has a name (Sum), its value is also continuously displayed in
the upper left of the window.
Notice that the lines l4, l5 and l6 have the property .display. This tells Geometer
to display their lengths in the window. We could have done this without using the
.display property simply by giving names to the numbers DE, DF and DG. In that case,
the named lines would no longer have the .display attribute and the final four lines
would look like this, and the view on the screen would remain the same:
DE = .f.vv(v4,
DF = .f.vv(v4,
DG = .f.vv(v4,
S = .f.rpn(DE,

v5,
v6,
v7,
DF,

"a");
"b");
"c");
DG, .add, .add, "Sum");

Once you have made the diagram and tested the conjecture by moving the point D
inside the triangle to verify that the sum does seem to remain constant, you can begin
to look for a proof.
Note: With some effort you could avoid editing the text file to present the sum.
Using standard techniques you could construct a line that is the sum of the three lengths
and display its length.
When you are trying to show something about a sum of lengths in a geometric way,
it is often a good idea to try to make a construction where the three lengths are set end
to end. In addition, in this case where you are trying to show that the sum is constant,
it would be nice to know what that constant might be.
The first approach seems difficult, but we can get some idea of what the sum might
be by looking at some special positions for D. These special positions might include
the center, points on the edge, or points at the vertices of the triangle 4ABC. If D is
very close to a vertex, almost the entire sum will be due to one length, and that length
is basically the altitude of the triangle.
Why would the sum be equal to an altitude? If it is true, might that lead to a proof?
One of the most common uses for the altitude of a triangle is to find its area which is
half the base times the altitude. If point D is inside the triangle, the three segments are
not altitudes of triangle 4ABC, but if we connect D to the vertices with line segments,
they are the altitudes of triangles 4ADB, 4BDC and 4CDA.
Since 4ABC is equilateral, all three of those smaller triangles have the same
length base, so the area of each is the half the product of that base times the altitude.
Since the three triangles combine to give the full triangle 4ABC, their areas must add
to the area of the original triangle.
That is basically the proof. Here is a more formal presentation:
We know that
A(4ABC) = A(4ADB) + A(4BDC) + A(4CDA).

(4.8)

If h is the altitude of 4ABC and s = AB = BC = CA, then equation 4.8

becomes:
s(DE) s(DF ) s(DG)
s(DE + DF + DG)
sh
=
+
+
=
,
2
2
2
2
2

4.7. FIND A RADIUS

89

so we conclude that DE + DF + DG = h.

4.7 Find a Radius

Given a right angle at A and two different
points B and C on one of the side of that angle, find the radius (in terms of the lengths of
the segments AB and AC) of the circle that
passes through B and C and is tangent to the
other side of the angle.
This is an easy problem, but if you decide
to investigate it with Geometer, you apparently immediately encounter a much harder
problem, since it is not particularly easy to
construct a line through two points and tangent to a line5 .

T
A

O
B

Figure 4.5: Tangent Circle Diagram

Proofaid/Radius.T [M]

To draw a useful diagram, however,

all you need is to draw something with
a circle tangent to one ray of a right anD
gle and passing through the other ray in
two points labeled B and C. To do so,
draw a right angle, draw a free point
A
B
O to be the center of the circle, drop a
perpendicular line from O to one side
of the angle, and the point of interFigure 4.6: Sum of Lengths
section will be the point of tangency.
Proofaid/EqualSum.T [M]
The intersections of the circle with the
other line are B and C. See figure 4.5.
(Alternatively, you can draw a circle, then a tangent to it, then a perpendicular to the
tangent passing through the circle at B and C.)
With the drawing, it is not hard to find the answer. Drop a perpendicular from O to
the other side of the angle at X. The point X is clearly the midpoint of segment BC,
and AX is clearly the same as OT , the radius of the circle that we are seeking. Clearly
AX = (AB + AC)/2.

4.8 Sum of Lengths

Show that if 4ABC is equilateral and point D lies on its circumcircle on the arc BC,
then AD = BD + CD. See figure 4.6.

4.8.1 Geometer Analysis of the Problem

5 It

can certainly be done; see Section 3.6.5.

90
It is quite easy to draw the Geometer diagram, but when we test it, we
seem to obtain almost no information.
If you do not believe that it is true, you
can help convince yourself by clicking on the segments AD, BD and CD
and issuing the Toggle Display Value
command which causes their lengths
to be displayed in the upper left portion of the Geometer window. The
three lengths will be displayed, but
you will need to do some mental arithmetic each time to convince yourself
that AD = BD + CD each time.

CHAPTER 4. COMPUTER-AIDED PROOF

C
C
D

Figure 4.7: Sum of Lengths

Proofaid/EqualSum.T [T]

You can add a few lines of code with the text editor to display not only the lengths
of the segments, but the result of calculating S = BD + CD AD. If you do that, the
lengths and their difference is constantly calculated and you will see that it is always
zero.
Here is the code to display these values, assuming the internal point names for A,
B, C and D are v1, v2, v3 and v4, respectively:
dbd = .f.vv(v2, v4, "BD");
dcd = .f.vv(v3, v4, "CD");
dad = .f.vv(v1, v4, "AD");
diff = .f.rpn(dbd, dcd, .add, dad, .sub, "BD + CD - AD");

To make some progress, why not construct a segment of length BD + CD? The
easiest way to do that is to extend the segment DB by an additional CD as in figure 4.7.
In that figure, C 0 is on the line BD and C 0 D = CD. We would like to show that
C 0 B = AD. It is probably hard to work with just a segment pointing out into the void,
so connect C 0 to C so that at least the new point will be part of a couple of triangles.
With this simple addition, the Test Diagram command returns a lot of information,
some of which is quite interesting:
Equal angles:
(A C D) (B C C)
Equal length segments:
[C D] [D C] [C C]

If we can show that these are true, we will know that 4ACD
= 4BCC 0 , and we
0
will be able to conclude that AD = BC which is what we need to show.
It looks like the little triangle 4CDC 0 is equilateral and if we can show that, it
will be sufficient to show that the triangles in the previous paragraph are congruent.
But 4ABC is equilateral, so clearly CDB = 120, so CDC 0 = 60 . Since we
constructed CD = C 0 D we can show that 4CDC 0 is equilateral.

4.9. BISECTOR BISECTOR

91

4.8.2 Proof of the Theorem

Extend segment BD to C 0 so that DC 0 = DC.
Since 4ABC is equilateral and ABDC is a cyclic quadrilateral we know that
60 = BAC is supplementary to BDC, so BDC = 120 . Since BDC and
CDC 0 together make a 180 angle, CDC 0 = 60 . But CDC = DC 0 , so 4CDC 0
is equilateral and thus CC 0 = CD.
Since 4ABC is also equilateral, AC = BC, and since ACD = BCD+60 =
BCC 0 we can conclude that 4ACB
= 4BCC 0 . Since the corresponding parts of
congruent triangles are congruent, AD = BC 0 = BD + CD.

4.9 Bisector Bisector

C

Show that if 4ABC is a right triangle

with C = 90 , then the angle bisector at C bisects the angle formed by the
median and the altitude from C.

We can easily construct a Geometer diagram where it is possible to disA

F E
O
B
play all possible right triangle shapes
by beginning with the hypotenuse AB
and choosing C to lie on the circle
Figure 4.8: Equilateral Projections
whose diameter is the hypotenuse AB.
Proofaid/Bisect.T [M]
If we then draw in the altitude AF , the
angle bisector AE and the median AO we obtain figure 4.8.

4.9.1 Geometer Analysis

If we test the diagram, we obtain quite a few relations, but among them are these:
Equal angles:
(F C E) (E C O)
Equal angles:
(A B C) (B C O) (A C F)
Equal length segments:
[A O] [O C]

Since CE is the angle bisector, if ACF = BCO we can subtract those from
the known equal angles ACE and BCE to obtain the result we want. The last
relation above reminds us that all the radii of a circle are equal, and that is plenty of
information to build a proof.

CHAPTER 4. COMPUTER-AIDED PROOF

92

4.9.2 Proof of the Result

Since CO is a median, point O is the midpoint of segment AB. We know that if
ACB = 90 then C must lie on a circle whose diameter is AB. Thus AO = BO =
CO. Since CF is an altitude, CF A = 90 and since F AC is equal to itself, by
AA, 4ACF 4ABC. Because of that similarity, ACF = ABC.

Since CO = BO, 4BOC is isosceles, so OBC = OCB. But BOC =

ACF , so ACF = OCB. Segment CE is the bisector of ACB, so ACE =
BCE. If we subtract the equal angles OCB and ACF from them, we obtain
F CE = ECO, which means that CE bisects F CO.

4.10 The Simson Line

Let M be any point on the circumcircle
of an arbitrary 4ABC. Let A0 , B 0 and
C 0 be the perpendicular projections of M
on the edges of 4ABC. (In other words,
drop perpendicular lines from M to each
of the edges of the triangle and find the
points of intersection. At least one of
those points has to lie outside the triangle.) Then A0 , B 0 and C 0 lie on a common line called the Simson line of the
triangle relative to the point M . See figure 4.9.

B
A

C
A
B

Figure 4.9: The Simson Line

Proofaid/Simson.T [M]

4.10.1 Geometer Analysis of the Simson Line

When the diagram above is tested, the following interesting relationships appear:
Points
C
Points
B
Points
A

on
M
on
M
on
M

a circle:
B A
a circle:
C A
a circle:
B C

That these relationships hold is obvious, since the perpendicular lines that were
dropped from M form right angles. But when sets of four points are concyclic, we
can find many equal angles, and a good way to show that the points A 0 , B 0 and C 0 are
collinear is to show that BA0 C 0 = CA0 B 0 . Since BC is a line, B 0 A0 C 0 must also
be a line since equal vertical angles are formed.

4.11. A PROPERTY OF TWO CEVIANS

The nice thing is that BA0 C 0 and

CA0 B 0 are both inscribed in the circles
with four concyclic points.

93

Since M is on the circumcircle of the

triangle, ABM C are concyclic points so we
can relate BM C to BAC. But we also
know that B 0 M C 0 A are concyclic so we can
relate C 0 M B 0 to the same BAC.
From these observations, the proof is
easy.

Figure 4.10: The Simson Line

Proofaid/Simson1.T [M]

4.10.2 Proof of the Existence of the Simson Line

The proof depends on the fact that M is on the arc CB (which will cause A0 to be
between B 0 and C 0 ). If this is not the case, simply relabel the figure so that it is.
Since M A0 BC and M C 0 AB right angles are formed at A0 and C 0 so
M A0 BC 0 are concyclic. Similar reasoning shows us that M A0 B 0 C and B 0 M C 0 A also
form concyclic sets of points. Draw the three circles containing those sets of points and
in addition, the line segments M B and M C. See figure 4.10.
Since the points M A0 BC 0 are concyclic, BA0 C 0 = BM C 0 . Similarly, since
the points M A0 B 0 C are concyclic, CA0 B 0 = CM B 0 .
Since C 0 M B 0 = C 0 M B + BM B 0 and BM C = BM B 0 + B 0 M C,
we will be done if we can show that C 0 M B 0 = BM C since this will imply that
C 0 M B = B 0 M C. Therefore BA0 C 0 = CA0 B 0 which will tell us that equal
vertical angles are formed and thus the points A0 , B 0 and C 0 are collinear.
Since M ABC form a concyclic set of points, BM C and BAC are supplementary. Since B 0 M C 0 A are concyclic, C 0 M B 0 and BAC are also supplementary.
Since BM C and C 0 M B 0 are supplementary to a common angle, they are equal,
and the proof is complete.

4.11 A Property of Two Cevians

In figure 4.11, let 4ABC be any triangle and choose D and E on the segments BC and
AB, respectively. Construct the two circles with diameters AD and CE that intersect
at points I and J. Show that IJ passes through the orthocenter H of 4ABC.

Notice that the line IJ will be the radical axis of the two circles. See Section 5.7
for information about the radical axis.

CHAPTER 4. COMPUTER-AIDED PROOF

94

4.11.1 Geometer Analysis of the Two-Cevian Problem

Since the conjecture involves the orthocenter H, it is only reasonable to draw the altitudes AA0 , BB 0 and CC 0 of the triangle before doing the test since the common
intersection of those altitude lines define the orthocenter. When the diagram is tested
by Geometer, there is not too much of interest, but there are a couple of things:
Points
H
Points
C
Points
A

on
I
on
E
on
D

a line:
J
a circle:
I J C
a circle:
I J A

We are reminded that the feet of the altitudes AA0 and CC 0 also lie on the circles.
This is obvious since the altitudes meet the triangle bases at 90 angles. There are no
useful ratiosonly the obvious ones, but we do notice that Geometer reports that H,
I and J lie on a line. This seems obvious, but remember, H was defined to be at the
intersection of a pair of altitudes, so Geometer does not know that it lies on IJ (and
neither do wethat is what we are trying to prove).
This does provide an idea. Draw the
segments HI and HJ and see what Geometer has to say. To show that the three points
lie on a line, we may be able to prove that
those new segments cause that force vertical angles to be equal or something similar.
When we do this, Geometer testing yields
the following interesting new relationships:
Ratios:
[A H] x [H A] = [C H] x [H C]
[B H] x [H B] = [H I] x [H J]
[A H] x [H A] = [H I] x [H J]

If these are true, then H does lie on IJ,

since IJ is the radical axis (see Section 5.7)
of the two circles and H is a point with
equal powers relative to those two circles
and hence lies on the radical axis.

B
E
I
C
A

A
D

H
B

C
J

Figure 4.11: Cevians as Diameters

Proofaid/Cevians.T [M]

4.11.2 Proof of the Two-Cevian Problem

Using again figure 4.11, we observe that A0 , the foot of the altitude from A, lies on the
circle with diameter AD since AA0 D is a right angle. Similarly, C 0 lies on the circle
with diameter CE.
If I and J are the points of intersection of the two circles, then IJ is the radical
axis of those two circles. Since it is the radical axis, we know that the intersection H 1

4.12. TANGENT LINE PROBLEM

95

of this axis with CC 0 will satisfy H1 C 0 H1 C = H1 I H1 J and the intersection of H2

with AA0 will satisfy H2 A0 H2 A = H2 I H2 J.
If we can show that H, the orthocenter, satisfies HA HA0 = HC HC 0 we are
done since H will be equal to both H1 and H2 .
Since AHC 0 = CHA0 and HC 0 A = HA0 C we know that 4AHC 0
4CHA0 by AA. This similarity tells us that HA HA0 = HC HC 0 holds and our
theorem about the two cevians is proved.

4.12 Tangent Line Problem

E
D

C
A

Figure 4.12: Tangent Line Problem

Proofaid/Zvezda.T [M]

If AB is the diameter of a circle choose X

anywhere on the segment AB. Construct
the two circles having AX and XB as diameters, and find the point E on the larger
circle that is the intersection of the circle
and the perpendicular to AB through X. If
AE and BE intersect the smaller circles at
C and D, show that CD is the common external tangent of the smaller circles6 . See
figure 4.12.

4.12.1 Geometer Analysis of the Tangent Line Problem

If we test the Geometer diagram for relations, we obtain the following possibly interesting relationships:
Points on a circle:
X E D C
Points on a circle:
A B D C
Right angles:
(B E A)

Since E is on the larger circle, it is obvious why BEA = 90 . The cyclic quadrilaterals are also likely to help, since they tell us a number of things about opposite
angles.
How does one show that a line is tangent to a circle? One way is this: If XCD =
XAC we are done, since they both cut off the same arc XC of the circle. Clearly we
need only show that CD is tangent at C, since a totally symmetric argument will show
that it is also tangent at D.

4.12.2 Proof of the Tangent Line Problem

6 Thanks

to Zvezdelina Stankova for showing me this problem..

CHAPTER 4. COMPUTER-AIDED PROOF

96

See figure 4.13. Construct segments CX,

DX and a perpendicular to AB through X
to a point Y on the line CD. Let =
CAX. If we can show that Y CX = ,
we will be done.

E
D

X
B
Since both cut off the same arc XC,
CAX = Y XC = . In addiFigure 4.13: Tangent Line Problem
tion to AEB, we know that ECX =

Proofaid/Zvezda1.T [M]
EDX = 90 so CEDX is a rectangle. The diagonals of a rectangle bisect each
other, so XY = CY . Thus 4CY X is isosceles, so Y CX = Y XC = , so CY
must be tangent to the circle with diameter AX.

4.13 Cyclic Quadrilateral Problem

In figure 4.14, let ABCD be a cyclic
quadrilateral. Draw four circles having each
of the sides of ABCD as a diameter. Show
that the intersections of adjacent sets of circles E, F , G and H also lie on a circle.

B
F
A

H
D

Figure 4.14: Cyclic Quadrilateral Problem

Proofaid/CyclicQuads.T [M]

4.13.1 Geometer Analysis of the Problem

A Geometer test of the diagram contains the following interesting relationships:
Points
B
Points
A

on
D
on
C

a line:
F H
a line:
E G

We know that the usual way to show that four points are concyclic is to show that
the opposite angles are supplementary, so draw those lines in addition to the lines AC
and BD (which appear to pass through E and G or F and H). To get some handle on
the interior angles it may be interesting to draw lines like DG so we do that and re-test
the diagram (figure 4.15) which yields more interesting information:
Parallel lines:

4.14. A STRANGE RELATIONSHIP

97

(AD) (EH)
Parallel lines:
(CD) (EF)
Parallel lines:
(BC) (FG)
Parallel lines:
(AB) (GH)
Perpendicular lines:
(AC) (DG)

We see that DG AC because they are both inscribed in a semicircle, and that
the same thing would hold for any of the other circle diameters.
The parallel lines show us that GAD = GEH, et cetera. If we can show that
GEH = GHF we will have solved the problem.
By now we have plenty of information to construct the proof.

4.13.2 Proof of the Cyclic Quadrilateral Problem

Since CGHD is a cyclic quadrilateral
with CD as diameter, DGC is a right
angle. Similarly, AGD is inscribed in
the circle with diameter AD, so AGD
is also a right angle. These two facts show
us that G is on the line AC. Similar arguments about the other points prove that
AEGC are all on a line and similarly for
BF HD.
Because they are both inscribed in
the same circle with diameter AD, we
know that GF D = GAD. Similarly,
CEH = CBH.

B
F
A

H
D

Figure 4.15: Cyclic Quadrilateral Problem

Proofaid/CyclicQuads1.T [M]

But CBD = CAD since they are

both inscribed on the original circle so we can conclude that GF H = GEH and
therefore EF GH is a cyclic quadrilateral.

4.14 A Strange Relationship

Let triangle 4ABC have its angle at A bisected by a segment AD, where D is on the
line BC. At D construct a perpendicular to AD intersecting AB at E. Find the length
of AE in terms of the lengths of AB and AC 7 .
7 Thanks

to Tatiana Shubin for this problem.

CHAPTER 4. COMPUTER-AIDED PROOF

98

What is surprising and perhaps a bit

strange is that the relationship sought
does not seem to depend on the angle at
Aonly on the lengths of the two sides
that meet at that angle. As usual, let us
call the lengths of the sides AC and AB
b and c, respectively.

C
D

E
B

Figure 4.16: A Strange Relationship

Proofaid/Strange.T [M]

4.14.1 A Geometer Analysis of the Problem

If we test the diagram as it stands, Geometer returns no interesting information. It does
tell us that AC/CD = AB/BD, but that is simply because AD is an angle bisector.
We will need to make some
construction to proceed. Notice
b
that if we can find the length of eix
ther AE or EB in terms of b and
c
we are done, so it seems that it
y C
might be a good idea to construct
G
D
a line from E that makes some
E
F w c-x
similar triangles that include AE
B
and/or EB as sides. The approach used here is to construct
a line through E and parallel to
Figure 4.17: A Strange Relationship
AC and intersecting CD at F . At
Proofaid/Strange.T [M]
least this will give us similar triangles 4ABC 4EBF that involve the unknown lengths.

With this we can show that EB/AB = EF/AC, but we do not know the length of
EF in terms of anything else, so it is still not clear what to do.
If we construct this segment EF and again use Geometer to test the diagram we
obtain these interesting observations:

Equal angles:
(A E D) (D E F)
Equal length segments:
[C D] [D F]

We would like to express the length of EF in terms of some other lengths, and if
we can show that CD is indeed equal to DF , perhaps that could be of use. In fact,
if we extend segments DE and AC to meet at G, we will have an isosceles triangle
4AGE, and it looks like CG = EF .
These hints provide enough information to find the proof.

4.15. SUM OF POWERS

99

4.14.2 Calculation of the Relationship

Label the lengths AB = c, AE = x (and therefore EB = c x), EF = w, CG = y
as shown in figure 4.17. Clearly 4AGE is isosceles since its altitude AD bisects the
angle at A. Since EF k AG, and those parallel lines are cut by transversal CF , we
know that DEF = DGC. We also know that CDG = F DE since they are
vertical angles. Thus 4DCG
= 4DF E so y = w.

Since y = w and b + y = AG = AE = x, w = x b. Since EF k AC,

4ABC 4EBF , so we know that AC/AB = EB/EF . This gives:
b
w
xb
=
=
.
c
cx
cx
Thus b(c x) = c(x b) and a little algebra gives us the result we need:
x=

2bc
.
b+c

There are of course other ways to do this. For example, you can compare the areas
of the triangles: A(4ABC) = A(4CAD) + A(4BAD), and the areas of each can
be expressed as A(4ABC) = bc sin(CAB)/2, et cetera. A brute-force calculation
using analytic geometry also works.

4.15  Sum of Powers

If M is a point on a circle, and ABCD is
a square inscribed in that circle, show that
M A2 +M B 2 +M C 2 +M D2 is independent
of the position of M . Suppose the square
is replacedP
by a regular n-gon A1 A2 . . . An .
Let Sm = i M Am
i . For what integers m is
Sm independent of the position of M ?

sum = 30.000
M

4.15.1 Geometer Analysis

Clearly, it does not matter which square or
n-gon is chosen, since we plan to rotate the
point M to all possible positions relative to Figure 4.18: Sum of Powers of Lengths
Proofaid/Pent4.T [M]
the polygon. Similarly, it does not matter
how big a circle is chosen, since if we multiply the size of the circle by a constant
, all of the lengths M Ai will be multiplied by the same , so the sum Sm will be
multiplied by the constant m . In case we decide to approach the proof later using analytic geometry, it is probably best to select a circle of radius 1 so that it will be easy to
express the vertices of the polygons in terms of sine and cosine without any extraneous
constants.

100

CHAPTER 4. COMPUTER-AIDED PROOF

Thus, our Geometer diagram might as well have one of the points of the n-gon at
(1, 0) and the others equally spaced around the circle. The coordinates of the n points
will be (cos((i/n)360), sin((i/n)360 )), where 0 i < n. Since we will want to
experiment with different polygons and different powers of m, the Geometer diagram
should be structured so that it is easy to modify. Here is a reasonable way to do so that
generates the diagram in figure 4.18 (where the polygon is a regular pentagon).
.geometry "version 0.51";
m = .f.rpn(4.000000);
N = .f.rpn(5.000000);
ang = .f.rpn(360.000000, N, .div);
ctr = .v.ff(0.000000, 0.000000, .in);
pt = .v.ff(1.000000, 0.000000, .in);
c1 = .c.vv(ctr, pt);
M = .vonc(c1, -0.461465, 0.887158, .red, "M");
x0 = .f.rpn(ang, 0.000000, .mul, .cos);
y0 = .f.rpn(ang, 0.000000, .mul, .sin);
v0 = .v.ff(x0, y0);
x1 = .f.rpn(ang, 1.000000, .mul, .cos);
y1 = .f.rpn(ang, 1.000000, .mul, .sin);
v1 = .v.ff(x1, y1);
x2 = .f.rpn(ang, 2.000000, .mul, .cos);
y2 = .f.rpn(ang, 2.000000, .mul, .sin);
v2 = .v.ff(x2, y2);
x3 = .f.rpn(ang, 3.000000, .mul, .cos);
y3 = .f.rpn(ang, 3.000000, .mul, .sin);
v3 = .v.ff(x3, y3);
x4 = .f.rpn(ang, 4.000000, .mul, .cos);
y4 = .f.rpn(ang, 4.000000, .mul, .sin);
v4 = .v.ff(x4, y4);
d0 = .f.vv(v0, M);
d1 = .f.vv(v1, M);
d2 = .f.vv(v2, M);
d3 = .f.vv(v3, M);
d4 = .f.vv(v4, M);
s = .f.rpn(d0, .log, m, .mul, .exp,
d1, .log, m, .mul, .exp,
d2, .log, m, .mul, .exp,
d3, .log, m, .mul, .exp,
d4, .log, m, .mul, .exp,
.add, .add, .add, .add, "sum");

The Geometer code works by calculating the x and y coordinates of each of the
regularly-spaced points and then creating those points. The values d0, d1, et cetera, are
the distances between each point and the variable point M , and the final calculation
of s raises each distance to the power m and adds them all up. Since the s value in
the Geometer diagram has a name, it is presented on the computer screen and updated
with each movement of the point M .
As we view the diagram in P
Geometer, we see that no matter where we move the
red point M , the value of S4 = i M A4i is displayed, and we see that it seems to have
the constant value 30.0000.
In this Geometer code, m is the exponent, set to 4 in this code fragment. We can
try different values of m simply by changing the second line. If m = 2, we find that the

4.15. SUM OF POWERS

101

sum is a constant 10.0000. If m = 6, the sum seems to be a constant 100.0000. In fact,

all even powers of m seem to yield constant values of the sum.
The following version of the code (found in Proofaid/Pent4m.T) is one line longer,
but it uses macro calls to generate points. It is easier to modify, however, than the
version above.
.geometry "version 0.60";
m = .f.rpn(4.000000);
N = .f.rpn(5.000000);
ang = .f.rpn(360.000000, N, .div);
ctr = .v.ff(0.000000, 0.000000, .in);
pt = .v.ff(1.000000, 0.000000, .in);
c1 = .c.vv(ctr, pt);
M = .vonc(c1, -0.434232, 0.900801, .red, "M");
.macro .vertex v(.flt f)
{
x = .f.rpn(ang, f, .mul, .cos);
y = .f.rpn(ang, f, .mul, .sin);
.return vv = .v.ff(x, y);
}
f0 = .f.rpn(0.000000);
f1 = .f.rpn(1.000000);
f2 = .f.rpn(2.000000);
f3 = .f.rpn(3.000000);
f4 = .f.rpn(4.000000);
v0 = v(f0);
v1 = v(f1);
v2 = v(f2);
v3 = v(f3);
v4 = v(f4);
d0 = .f.vv(v0, M);
d1 = .f.vv(v1, M);
d2 = .f.vv(v2, M);
d3 = .f.vv(v3, M);
d4 = .f.vv(v4, M);
s = .f.rpn(d0, .log, m, .mul, .exp,
d1, .log, m, .mul, .exp,
d2, .log, m, .mul, .exp,
d3, .log, m, .mul, .exp,
d4, .log, m, .mul, .exp,
.add, .add, .add, .add, "sum");

How about different regular polygons? We will construct a file by modifying the
macro-based file we used for the pentagon. The value N in the third line is the number
of vertices, so if we would like to look at a regular heptagon (7-sided figure), we just
change the value of N to 7. We also need to add two vertices, v5 and v6, two more
distances, d5 and v6, and to modify the formula for the sum at the end of the file.
Here is the Geometer code for a regular heptagon (found in Proofaid/Hept7m.T,
and we find that for m = 4, the result seems to be a constant 42.0000. Again, we find
that different even values of m generate constant sums and odd values do not. On the
CD you can also find a file Proofaid/hept7.T which does the same thing, but is based
on the non-macro version Proofaid/pent4.T.

102

CHAPTER 4. COMPUTER-AIDED PROOF

.geometry "version 0.60";

m = .f.rpn(4.000000);
N = .f.rpn(7.000000);
ang = .f.rpn(360.000000, N, .div);
ctr = .v.ff(0.000000, 0.000000, .in);
pt = .v.ff(1.000000, 0.000000, .in);
c1 = .c.vv(ctr, pt);
M = .vonc(c1, -0.603538, 0.797334, .red, "M");
.macro .vertex v(.flt f)
{
x = .f.rpn(ang, f, .mul, .cos);
y = .f.rpn(ang, f, .mul, .sin);
.return vv = .v.ff(x, y);
}
f0 = .f.rpn(0.000000);
f1 = .f.rpn(1.000000);
f2 = .f.rpn(2.000000);
f3 = .f.rpn(3.000000);
f4 = .f.rpn(4.000000);
f5 = .f.rpn(5.000000);
f6 = .f.rpn(6.000000);
v0 = v(f0);
v1 = v(f1);
v2 = v(f2);
v3 = v(f3);
v4 = v(f4);
v5 = v(f5);
v6 = v(f6);
d0 = .f.vv(v0, M);
d1 = .f.vv(v1, M);
d2 = .f.vv(v2, M);
d3 = .f.vv(v3, M);
d4 = .f.vv(v4, M);
d5 = .f.vv(v5, M);
d6 = .f.vv(v6, M);
s = .f.rpn(d0, .log, m, .mul, .exp,
d1, .log, m, .mul, .exp,
d2, .log, m, .mul, .exp,
d3, .log, m, .mul, .exp,
d4, .log, m, .mul, .exp,
d5, .log, m, .mul, .exp,
d6, .log, m, .mul, .exp,
.add, .add, .add, .add, .add,
.add, "sum");

By making modifications like those above to the Geometer file, we can construct
a table of values for various regular polygons and various even exponents. In the table
below, a is placed when the sum is not constant.
Notice that the column entries for m = 0 are trivial to calculate, so they are included as well. Finally, what seems to be important is not that the inscribed figure is a
polygon but rather that it consists of n equally-spaced points, we can add the data for
n = 1 and n = 2, and both of these observations make the pattern clearer.

4.15. SUM OF POWERS

103

n\m 0
2
4
6
8
10
12 14
1 1

2 2
4

3 3
6 18

4 4
8 24
80

5 5 10 30 100 350

6 6 12 36 120 420 1512

7 7 14 42 140 490 1764 6468

Each row appears to consist of the following multiples of n:
1, 2, 6, 20, 70, 252
and 924, when they are constant. These are just the values of 2k
k in Pascals triangle,
so the general formula appears to be:


n1
X
m
M Am
=
n
,
Sm =
i
m/2
i=0
where the formula works for all even values of m less than 2n.

4.15.2 Proof of the Result

For the case where m = 2, it is easy to prove the result for an arbitrary regular polygon
with n sides. On the unit circle, the vertices have coordinates (cos k , sin k ), where
0 k < n and k = (k/n)360 .
Let the point M have coordinates (cos , sin ). Then the sum S2 is given by:
S2

n1
X
k=0

(cos k cos )2 + (sin k sin )2

(4.9)

If we multiply out the right hand side of equation 4.9 and simplify the result, we
obtain:
S2

n1
X

k=0
n1
X

k=0
n1
X
k=0

cos2 k 2 cos k cos + cos2 + sin2 k 2 sin k sin + sin2

cos2 k + sin2 k 2(cos k cos + sin k sin ) + cos2 + sin2
2 2(cos k cos + sin k sin )

n1
n1


X
X
= 2n 2 cos
cos k + sin
sin k .
k=0

If we can show that the sums

If we recall that
cos =

k=0

Pn1
k=0

ei + ei
2

cos k and
and

Pn1
k=0

sin k are zero, we are done.

sin =

ei ei
2i

CHAPTER 4. COMPUTER-AIDED PROOF

104

then the sums that interest us become a pair of geometric series. For example:
n1
X

cos k =

k=0

1  X (k/n)360 X (k/n)360 
e
e
+
.
2
n1

n1

k=0

k=0

(4.10)

Both series on the right hand side of equation 4.10 begin with 1 and have a ratio of

either e(1/n)360 or e(1/n)360 . Both series look like 1 + r + . . . + r n1 = (1

n
r )/(1 r), and in both cases r n = 1, so both sums are zero. The same thing happens
for the sum of the sines. Therefore the sum S2 is independent of .

4.15.3 Alternative Proof with Vectors

The same result can be obtained with vectors. Let Ak for 0 k < n be the vertices
of a regular n-gon inscribed in a circle with center O. If M is any point on the circle,

then M Ak =M O + OAk . Then:

S2 =

n1
X

(M A i )2 =

k=0

n1
X

(M O + OAi )2 .

k=0

Expanding the expression on the right, we obtain:

S2

n1
X

k=0
n1
X

MO +

k=0

n1
X
k=0

2 MO

= n M O +2 M O
2

(M O +2 M O OAi + OAi )

n1
X

OAi

OAi

k=0

n1
X

OAi

k=0
n1
X 2
+
OAi
k=0

Since M O and Ak are both constant and equal to r 2 , where r is the radius of the
circle, this is equivalent to:
S2
Pn1
k=0 OAi = 0 ,

= nr + nr + 2 M O

n1
X

OAi

k=0

Clearly,
since if it did point in any direction, by symmetry, it would
have to point the same amount in every one of n equally-spaced directions. Thus the
sum S2 = 2nr2 which is what we obtained in the previous section with r = 1.

Chapter

More Useful Theorems

The results in this chapter are generally not covered in high school geometry courses,
but they are all very useful tools for solving other geometric problems. In this chapter
we will continue to use computer techniques to solve problems, so if you have not
already done so, it may be useful to read Chapter 4 before continuing.

5.1 Cevas Theorem

Cevas (pronounced chevas) theorem is perhaps the most useful geometric theorem
that is not taught in a standard high school course.
Theorem 5.1 (Cevas Theorem) Given any triangle 4ABC with a point P a inside
it1 (see figure 5.1), if the lines from A, B and C through P intersect the opposite sides
1 In fact, it is not difficult to show that P can be outside the triangle as well, as long as it does not lie on
any of the (infinite) lines that make up the edges of the triangle.

105

CHAPTER 5. MORE USEFUL THEOREMS

106

of the triangle at points X, Y and Z, respectively, then:

AZ BX CY

= 1.
BZ CX AY

The converse is also true (although we

do not prove it here): If three lines emanate
from the vertices of a triangle and intersect
the opposite sides such that the ratios of the
lengths into which the sides are divided satisfy equation 5.1 then the three lines are concurrent. Cevas theorem is also true if the
ratios in equation 5.1 are directed ratios (see
Section 5.2.1).

C
Y
P

(5.1)

Figure 5.1: Cevas Theorem

This is a fairly amazing theorem, since

it applies to any triangle and to almost any
point. If you have trouble proving this theorem, you may find it instructive before reading on to prove it in certain special cases,
such as when P is the centroid, incenter, circumcenter, or orthocenter.
Bigtheorems/Ceva.T [M]

Any line from the vertex of a triangle through a point on the opposite side of that
triangle is called a cevian, so Cevas theorem refers to three cevians that happen to
pass through the same point.

5.1.1 Geometer Analysis of Cevas Theorem

If we make no progress on the proof, we can always draw a Geometer diagram and see
what relationships hold when we test that diagram. After moving around the points,
the only relationships that Geometer notices are a large number of triangle area ratios
that seem to hold:
Triangle
[A Z
[A Z
[A Z
[A Z
[A Z
[A B
[A B
[A B
[A B
[A B
[A B
[A B
[A B
[A B
[A B
[A B
[A B
[A B

area
P] x
P] x
C] x
C] x
C] x
P] x
P] x
Y] x
Y] x
X] x
X] x
X] x
C] x
C] x
C] x
C] x
C] x
C] x

ratios:
[A C P]
[A Y P]
[C X A]
[A C P]
[B Z P]
[A C P]
[C Y P]
[C Y P]
[B C P]
[C X P]
[A C P]
[C Y P]
[C X P]
[C Y P]
[B Z P]
[A Z P]
[A Y P]
[B X P]

=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=

[B
[B
[B
[B
[A
[B
[B
[B
[A
[B
[A
[A
[B
[B
[A
[A
[A
[A

Z
Z
Z
Z
Z
X
C
C
B
X
B
B
C
C
B
B
B
B

P]
C]
P]
C]
P]
P]
P]
Y]
P]
P]
P]
Y]
P]
Y]
P]
P]
Y]
X]

x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x

[B
[B
[B
[B
[B
[C
[A
[A
[B
[C
[C
[C
[C
[A
[B
[A
[A
[B

C
C
X
C
Z
X
Y
Y
C
X
X
X
X
C
Z
Z
C
C

P]
Y]
P]
P]
C]
P]
P]
P]
Y]
A]
A]
P]
A]
P]
C]
C]
P]
P]

5.1. CEVAS THEOREM

107

After examining a few, we notice that they all are based on the same idea. Relationships like that shown in the fifth line are the easiest to see:
[A Z C] x [B Z P] = [A Z P] x [B Z C]

This basically says that 4AZC and 4BZC have the same ratio as 4AZP and
4BZP . This is obvious because the ratios of the areas are the same as the ratio into
which Z divides the base AB. Once we see this, it is obvious why the first ratio in the
list is true: 4AZC is composed of the two triangles 4ACP and 4AZP and 4BZC
can be similarly decomposed.
Thus the ratios of the subdivisons of the sides are the same as the ratio of the areas
of the triangles 4ACP , 4BCP and 4ABP . Since Cevas theorem concerns these
ratios, we are probably on the right track. Using this information the proof is simple to
construct.

5.1.2 Proof of Cevas Theorem

Here is a proof, derived from the observations above:
Relative to the base AB, the triangles 4AZP and 4BZP have the same altitude,
their areas are proportional to AZ : BZ. The areas of triangles 4AZC and 4AZC
are similarly related so we obtain:
AZ
BZ

A(4AZP )
A(4BZP )

A(4AZC)
A(4BZC)

(5.2)

Triangle 4AZC is composed of triangles 4AZP and 4ACP and 4BZC can
be similarly decomposed, so their areas satisfy:
A(4AZC) = A(4AZP ) + A(4ACP )
A(4BZC) = A(4BZP ) + A(4BCP ).

(5.3)
(5.4)

Rearranging equations 5.3 and 5.4, we obtain:

A(4ACP )
A(4BCP )

= A(4AZC) A(4AZP )
= A(4BZC) A(4BZP ).

(5.5)
(5.6)

Combining equations 5.2, 5.5 and 5.6 by using the fact that if a/b = c/d = e where
b and d are different then (a c)/(b d) = e, we obtain:
AZ
BZ

(4ACP )
.
(4BCP )

(5.7)

There is nothing special about the selection of AB as a base of the triangle, and
similar calculations give us:
BX
CX
CY
AY

=
=

(4ABP )
(4ACP )
(4BCP )
.
(4ABP )

(5.8)
(5.9)

CHAPTER 5. MORE USEFUL THEOREMS

108

Now multiply together equations 5.7, 5.8 and 5.9 to obtain Cevas theorem:
(4ACP ) (4ABP ) (4BCP )
AZ BX CY

= 1.
BZ CX AY
(4BCP ) (4ACP ) (4ABP )

(5.10)

It is a good exercise to show how this proof can be modified to show that Cevas
theorem still holds if P is exterior to the triangle 4ABC.

5.2 Menelaus Theorem

Menelaus theorem is similar to Cevas theorem (see Section 5.1) but with the roles of
points and lines reversed.

Z
C

D C
Y

Y
B

Figure 5.2: Menelaus Theorem

Bigtheorems/Menelaus.T [M]

The theorem is illustrated in the diagram on the left of figure 5.1.

Theorem 5.2 (Menelauss Theorem) Let 4ABC be an arbitrary triangle. If a line
L that does not pass through A, B or C passes through the sides AB, BC and CA at
X, Y and Z, respectively, we have:
AX BY CZ

= 1.
BX CY AZ

(5.11)

We will not show it here, but like Cevas theorem, the converse of Menelaus theorem is also true: Given a triangle and points on the edges of that triangle whose
distances satisfy the equation 5.11, then those three points lie on a line.

5.2.1  Directed Ratios

Notice that no matter how the line L is drawn, it must intersect at least one of the
triangle segments (and sometimes all three of them, but never exactly two) outside the
triangle. For this reason Menelaus theorem is often stated with the product of the
ratios in equation 5.11 equal to 1, where the ratios are assumed to be directed ratios.

5.2. MENELAUS THEOREM

109

A directed ratio of two segments on the same line has the numeric value of the ratio
of their lengths, but is positive if they are directed in opposite directions and negative if
they are directed in the same direction. In figure 5.2, the directed ratios AX/BX and
BY /CY would be positive, but CZ/AZ would be negative.
If all three intersection points of the line L were outside the triangle, all three
directed ratios would be negative, again yielding a negative product of ratios.
Notice that Cevas theorem is also true when the ratios are assumed to be directed:
if P is outside the triangle, then two of the directed ratios in equation 5.1 will be
negative and one positive, so the statement of Cevas theorem is correct with +1 as the
product of the directed ratios.

5.2.2 Geometer Analysis of Menelaus Theorem

If the diagram on the left of figure 5.2 is subjected to the Geometers test for relationships, nothing is found. This is not too surprising, since there are so few constraints on
the geometry. When this happens, it is likely that the addition of one or more auxiliary
lines will help, so a good strategy is to add likely lines and rerun Geometers test.
Recalling the proof of Cevas theorem, a first approach might be to draw the segments CX, AY and BZ. If you do this, the test will discover a list of triangle area
relationships similar to those encountered when we tested the Ceva diagram, but although the set looks promising, it does not seem to yield enough useful ratios to solve
the problem.
When ratios play a big part in the statement of a theorem, it is often a good idea
to look for similar triangles since every pair of similar triangles in a figure can generate a whole list of preserved ratios. There are no similar triangles in the diagram
on the left of figure 5.2, but if we construct the line through C parallel to AB that
intersects line XY at D, suddenly we have a few. In particular, 4CDZ 4AXZ
and 4BXY 4CDY . These are nice, since those triangles seem to contain all
the lengths of segments that appear in the statement of Menelaus theorem in equation 5.11. There is nothing special about drawing the line through C parallel to the
opposite side; a similar line could have been drawn through A or B.
Testing the Geometer diagram with the added segment DC yields, as expected, an
interesting set of ratios that are preserved:
Ratios:
[A Z]
[A C]
[A C]
[B Y]
[B C]
[B C]
[B X]
[B X]
[A X]
[A X]

x
x
x
x
x
x
x
x
x
x

[Z
[Z
[X
[Y
[Y
[X
[Y
[C
[Z
[C

D]
D]
Z]
D]
D]
Y]
D]
Y]
D]
Z]

=
=
=
=
=
=
=
=
=
=

[C
[C
[A
[C
[C
[B
[X
[B
[X
[A

Z]
Z]
Z]
Y]
Y]
Y]
Y]
Y]
Z]
Z]

x
x
x
x
x
x
x
x
x
x

[X
[X
[X
[X
[X
[X
[C
[C
[C
[C

Z]
D]
D]
Y]
D]
D]
D]
D]
D]
D]

CHAPTER 5. MORE USEFUL THEOREMS

110

A quick examination shows us that all these ratios can be derived from the pair
of similar triangles listed in the previous paragraph. From here it is relatively easy to
construct a proof of Menelaus theorem by fiddling around with the ratios above.

5.2.3 Proof of Menelaus Theorem

Construct the line parallel to AB through C meeting line XY at D as in figure 5.2.
Since CD k AB, it is easy to show that 4CDZ 4AXZ and 4BXY 4CDY .
Since those triangles are similar, we have:
CZ
AZ
BY
CY
AX
CD
CD
BX

=
=
=
=

ZD
XZ
XY
YD
XZ
ZD
YD
XY

(5.12)
(5.13)
(5.14)
(5.15)

If we multiply together equations 5.12 through 5.15, we obtain:

CZ BY AX CD
ZD XY XZ Y D

,
AZ CY CD BX
XZ Y D ZD XY
which reduces to equation 5.11 and proves the theorem.

5.3 Alternate Proofs of Menelaus and Ceva

H
F
A
I

L
P

G
B

A
B K

C
C

Figure 5.3: Menelaus and Cevas Theorems

Bigtheorems/Ceva1.T, Bigtheorems/Menelaus1.T [P]

When clever constructions are made, the proofs of Cevas and Menelaus theorems
are easier to find. Figure 5.3 shows two of those constructions.
On the left, to prove Cevas theorem, construct a line through point A that is parallel
to BC and extend the segments BF and CG to meet that new line. For a quicker proof
of Menelaus theorem, use the diagram on the right. If line KLM cuts the three sides

5.4. USING MENELAUS AND CEVAS THEOREM

111

of the triangle, construct perpendiculars from A, B and C to line KLM meeting it

atA0 , B 0 and C 0 , respectively.
Try to find alternate proofs of Menelaus and Cevas theorems based on these new
constructions. Those proofs are not difficult, but the solutions can be found in the files
Ceva1.T and Menelaus1.T

5.4 Using Menelaus and Cevas Theorem

If you would like to practice using Cevas theorem, Chapter 7 on triangle centers contains a surprising number of examples where Cevas theorem leads to a proof of the
existence of some center.
If you would like to practice with Menelaus theorem, look in Chapter 9 on projective geometry where Menelaus theorem is used to prove the Euclidean version of a
number of projective geometry theorems.

5.5 Ptolemys Theorem

Ptolemys theorem provides a powerful method for calculating lengths in a quadrilateral whose vertices happen to be concyclic. Conversely, if you know the lengths of the
sides and diagonals, it can be used to determine if the quadrilateral is concyclic.

C
D

Figure 5.4: Ptolemys Theorem

Bigtheorems/Ptolemy.T [M]

The theorem is illustrated in the diagram on the left of figure 5.4.

Theorem 5.3 (Ptolemys Theorem) If A, B, C and D are four concyclic points in
clockwise or counterclockwise order on the circle, then the following relationship holds
and conversely:
AB CD + BC AD = AC BD.
(5.16)
Notice that Ptolemys theorem is a generalization of the pythagorean theorem; if the
quadrilateral ABCD is a rectangle, Ptolemys theorem is equivalent to the pythagorean
theorem.

CHAPTER 5. MORE USEFUL THEOREMS

112

5.5.1 Geometer Analysis of Ptolemys Theorem

Ptolemys theorem is not easy to prove, and if we simply draw the basic figure seen
on the left of figure 5.4 and test it with Geometers relationship finding commands, it
only tells us about a few obvious angles that are equal (since they are inscribed in the
same circle) or supplementary (since they are opposite angles in a cyclic quadrilateral).
It seems clear that we will have to use ratios and therefore similar triangles to make
any progress, so we need to convert equation 5.16 into ratios, but what are we going to
do about that annoying + sign?
You probably cannot avoid looking at the answer, since it is drawn on the right of
the figure, but here is why it (or any of the 7 other equivalent constructions) is a good
idea. To convert the product form of equation 5.4 into a ratio form, we need terms like,
for example, AB/AC. Thus we would like to get AC and AD to be corresponding
parts of similar triangles. If we imagine rotating 4ABD until its angle at A matches
a new triangle, it is clear that a rotation by CAB will align AB and AC. Of course
this will rotate AD out of the circle by the same angle CAB. Therefore, a reasonable
way to begin is to construct a line from A making angle CAB with the line AD, as
shown in the right side of figure 5.4. The new line must intersect the line CD at some
point H outside the circle.
Now when Geometer tests the diagram, we obtain a far more interesting set of
relations, including:
Equal angles:
(B C A) (C
Ratios:
[C H] x [A
[B C] x [A
[A B] x [A
[A B] x [C
[A B] x [D

H A) (A D B)
D]
H]
H]
H]
H]

=
=
=
=
=

[B
[D
[A
[A
[B

D]
H]
D]
C]
C]

x
x
x
x
x

[A
[A
[A
[B
[A

H]
C]
C]
D]
D]

The equal angle information is encouraging, since it shows clearly that 4ABD
4ACH, and the ratios, if we can show that they hold due to similarities of triangles,
include almost all of the lengths that appear in equation 5.4. We can also see from
the ratios that 4ABC must be similar to 4ADH. At first it is a bit annoying to see
lengths like CH and DH in the list, until we realize that CH = CD + DH! This
pulls in the missing CD, and gets a + sign into the equations.
Using these observations, the theorem is easy to prove.

5.5.2 Proof of Ptolemys Theorem

This is just a proof of one direction of the theorem, namely, that if the points are
concyclic, the relationship in equation 5.16 holds.
Construct a line from A making angle CAB with line AD that intersects the line
CD at H.

5.6. HERONS FORMULA

113

We know that ABD = ACD since they are inscribed in the same circle, and
DAB = HAC since they are both equal to DAC + CAB. Therefore by AA
similarity, 4DAB 4HAC.

We know that ABC and ADC are supplementary since they are opposite angles in an inscribed quadrilateral, and it is obvious that ADC and AHD are also
supplementary, so ADH = ABC. This, and the fact that BAC = DAH
allows us to conclude by AA similarity that 4ADH 4ABC.
We know that:

CD

DH

CH

(5.17)

and the two sets of similar triangles show us that:

BD
CH
AD
AB

=
=

AB
AC
DH
BC

=
=

AD
AH
AH
AC

(5.18)
(5.19)

To get rid of the DH and CH in equation 5.17, we can substitute their values
obtained from equations 5.18 and 5.19:
CD +

AC BD
BC AD
=
AB
AB

(5.20)

If we multiply equation 5.20 by AB, we obtain:

AB CD + BC AD = AC BD,
which proves the theorem.

5.5.3  Additional Notes on Ptolemys Theorem

It is possible to prove the theorem using brute force by listing the lengths of all six sides
and relating them with the law of sines. Using the fact that various sets of angles add
to 180 (because they form a triangle or are opposite angles in the cyclic quadrilateral),
we can expand the equations using the formulas for the sine and cosine of a sum of
angles. After about ten lines of calculation, Ptolemys formula drops out.
There is also a beautiful derivation of Ptolemys theorem based on inversion in a
circle. See Section 8.8.

5.6 Herons Formula

Herons formula provides an easy way to calculate the area of an arbitrary triangle
given the lengths of its sides:

CHAPTER 5. MORE USEFUL THEOREMS

114

Theorem 5.4 (Herons Formula) Given an arbitrary triangle 4ABC, let a, b and c
be the lengths of the sides opposite the vertices A, B and C, respectively. Let s =
(a + b + c)/2 be the semiperimeter of 4ABC. Then the area of the triangle is given
by the formula:
p
A(4ABC) = s(s a)(s b)(s c),
(5.21)

There is a straightforward but fairly messy trigonometric proof. Begin with the
observation that A(4ABC) = (ab sin )/2, where = ACB. We also know, from
the law of cosines, that c2 = a2 + b2 2bc cos . Solve these equations for sin and
cos , and using the fact that sin2 + cos2 = 1, do a lot of algebra, and you will
eventually arrive at the result.

5.6.1 Geometer Analysis of Herons Formula

O
A

Figure 5.5: Herons Theorem

Bigtheorems/Heron.T [M]

A more geometric proof would be nice, but it is a bit difficult to find the correct
figure. Recall that Section A.6.8 contains a lot of interesting formulas relating expressions like s, s a, s b and s c to the area of a triangle. This observation suggests
that perhaps it would be useful to consider a diagram that includes some combination
of incircles and excircles. The diagram that does the trick is shown in figure 5.5. The
lines OD and O 0 E are perpendiculars dropped from the centers of the circles to the
line AC.
If we use Geometer to test that diagram, all sorts of relationships appear, but the
ones that are most interesting are these:
Right angles:
(O C O) (A E O) (A D O)
Equal angles:
(A C O) (B C O) (E O C)
Ratios:
[C E] x [C O] = [O D] x [C O]
[C D] x [C O] = [O E] x [C O]
[C D] x [C E] = [O D] x [O E]
[A D] x [O E] = [A E] x [O D]

5.6. HERONS FORMULA

115

Since ACO = EO 0 C and OD and O 0 E form right angles with AC, 4COD
4O0 CE. Using the notation from Section A.6.8, we have DC = s c, CE = s b,
OD = r and O0 E = ra . This, plus a little algebra with the ratios is enough to yield
the result.

5.6.2 Proof of Herons Formula

If we can show that OCO 0 = 90 then we will know that ACO = EO 0 C since
both ACO and EO 0 C are complementary to O 0 CE. But DCB + BCE =
180 and those angles are bisected by CO and CO 0 so OCO0 = 90 .
Thus we know by AA that 4COD 4O 0 CE. From this similarity, we have:
CE
OD
= 0
DC
OE

or

r
sb
=
.
sc
ra

(5.22)

We also know from Section A.6.8 that:

A(4ABC) = ra (s a) = rs.

(5.23)

From equation 5.23 we see that r = A(4ABC)/s and ra = A(4ABC)/(s a).

If we substitute these values into equation 5.22, we obtain:
(s b)(s a)
A(4ABC)
=
s(s c)
A(4ABC)
Cross-multiply to obtain:
(A(4ABC))2 = s(s a)(s b)(s c),
which is equivalent to Herons formula.
Note: In a sense, the pythagorean theorem is a special case of Herons formula.
If c is the hypotenuse of a right triangle, then the area of that triangle is ab/2. If we
substitute s = (a + b + c)/2 into Herons formula, expand, and set it equal to ab/2, we
obtain:
p
ab/2 = (1/4) (2ab)2 (a2 + b2 c2 )2
a 2 b2
a 2 b2
(a2 + b2 c2 )2
=

4
4
16
0 = a 2 + b2 c2 .

5.6.3 Brahmaguptas Formula

Herons formula looks a lot like another interesting formula called Brahmaguptas formula that applies to cyclic quadrilaterals. If a quadrilateral is inscribed in a circle and

CHAPTER 5. MORE USEFUL THEOREMS

116

has sides of lengths a, b, c and d, and if s = (a + b + c + d)/2 is its semiperimeter,

then the area A of that quadrilateral is given by the formula:
p
A = (s a)(s b)(s c)(s d).
We will not prove Brahmaguptas formula here, but it can be proved relatively
easily with some algebra and the laws of sines and cosines. It can also be proved using
Herons formula on the two triangles formed if you extend two sides of the quadrilateral
to meet outside the circle. Herons formula is a special case of Brahmaguptas formula
where two of the points coincide, so effectively d = 0.

5.7 The Radical Axis

In this section we will make more sophisticated use of Geometers features. We will
begin by investigating the following question:
Given two circles, what is the set of points whose tangent lines to those two circles
have equal length?

(R2 + L2)1/2

T
R
O

Figure 5.6: pythagorean Theorem

Bigtheorems/Rat1.D [M]

Assuming we do not have a clue what the answer is, we can construct a Geometer
diagram that will give us a good idea. The diagram is based on the idea presented in
figure 5.6. If P is a point outside a circle of radius
R whose tangent distance to the
circle is L, then the length of the segment P O is R2 + L2 . Thus, to find all points
P with a tangent length of L to the circle, we need only construct any right triangle
like the one in the figure and then draw a circle about O having the hypotenuse of that
triangle as its radius.
To find the points satisfying our condition, we will build a diagram with the two
circles but which in addition contains a pair of points the distance between which will
be the L. We will copy this distance L on a tangent line to each circle, forming triangles
that correspond to 4P OT in figure 5.6. The circles with the hypotenuses of those

5.7. THE RADICAL AXIS

117

t
T O

L
Figure 5.7: Radical Axis Diagram
Bigtheorems/Radical1.T [M]

triangles as radii will represent all points whose tangent distance to the circles is L.
The points that interest us are at the intersections of those circles.
Figure 5.7 shows the result of this construction. The sizes and positions of the
circles can be changed by adjusting O, o, T and t. The tangent lengths are equal:
T P = tp = L. Two circles are drawn centered at O and o passing through P and p,
respectively. The points of intersection of these last two circles are points without a
name.
O
T

T O

t
OTo

T
O
o

t
T

o
t

O o

t
L

Figure 5.8: The Radical Axis

Bigtheorems/RadicalTest.T [M]

To eliminate much of the clutter for our experiments, change the colors of all the
circles except for the original two and of the tangent lines and the points P and p to

CHAPTER 5. MORE USEFUL THEOREMS

118

be the invisible color. We also change the colors of the points of intersection to be the
smear color. Then we move the endpoints of the line segment L and see what happens.
We can do this repeatedly with different arrangements of the two original circles to get
an idea of what sorts of possibilities occur. The results of six such experiments are
illustrated in figure 5.8.
From our experiments, it appears that in every case the locus of points satisfying
our conditions is a line or at least a part of a line which will be called the radical axis
of the two circles. It is not too difficult to extend the definition so that the radical axis
will be an entire line even if the two circles intersect. The problem with the definition
above is that when P is inside a circle, it is impossible to draw a tangent line.
If we examine figures A.29 and A.30 we see that if a secant of the circle through
P intersects the circle at points A and B, then P A P B = P T 2 . Instead of looking
for locations of P that make P T constant (or equivalently, make P T 2 constant), why
not find P such that P A P B is constant? Then we can use any secant we want. This
product, P A P B, is called the power of a point P relative to the circle. The radical
axis of two circles, then, is best defined as the set of points P that have equal powers
relative to both circles.

5.7.1 Proof that the Radical Axis is a Line

So far all we have is a lot of evidence obtained by playing with Geometer diagrams
that the points satisfying our conditions lie on a line. It is not difficult, now that we
know the answer, to construct a proof of that fact.

D
L2
L1
h1
T1

T1
R1

O1

R1
d1

T2
L

L
D

T2

h2

R2
R2

d2

O2

Figure 5.9: Radical Axis Proof

Bigtheorems/RadicalProof.T [M]

It seems that the easiest proof is algebraic. We will prove it only for the case that
the circles are disjoint as in figure 5.9. A similar proof can be written when the circles

5.7. THE RADICAL AXIS

119

intersect. We will show that the radical axis is perpendicular to the line connecting the
centers of the circles. Pick D on that line of centers such that DT1 = DT2 . We need to
show that for any point D 0 on the line perpendicular to O1 O2 through D that L1 = L2 ,
with the labeling as in the figure. To complete the proof, simply apply the pythagorean
theorem a few times and combine the results:
d21

= R12 + L2

(5.24)

d22
h21
h22
h21
h22

R22 + L2
d21 + h2
d22 + h2
R12 + L21
R22 + L22

(5.25)

=
=
=
=
=

(5.26)
(5.27)
(5.28)
(5.29)

Combining equations 5.24, 5.26 and 5.28 and similarly equations 5.25, 5.27 and
5.29we obtain:
L21
L22

= h2 + L 2
= h2 + L 2

which shows us that L1 = L2 , so the proof is complete.

5.7.2 The Radical Center

Notice that if we have a set of three circles whose centers are not collinear, then each
pair will have a radical axis. The point P of intersection of any two of the radical axes
will have the same power relative to all three circles, so it must lie on the third radical
axis. In other words, the three radical axes meet at a point called the radical center of
the three circles.

120

CHAPTER 5. MORE USEFUL THEOREMS

Chapter

Locus of Points
Computer geometry programs are particularly useful when studying locus problems
whose goal is to find all the points that satisfy certain conditions. The term satisfy
certain conditions can be very general, and with Geometer or a similar program, it is
often easy to generate hundreds of such points automatically. This visual display of a
large set of solution points often makes it quite easy to guess the answer, even if you
have no idea how to prove it.
Here we will examine a number of applications of computer geometry programs to
locus problems.

6.1 Unknown Geometric Locus



Many geometric problems are stated directly as locus problems: Given a circle with
points B and C fixed upon it, if the point A is allowed to move on the arc BC, what is
the locus of points of the incenter of 4ABC?
121

CHAPTER 6. LOCUS OF POINTS

122

With Geometer, it is easy to find out (or

at least to make a good guess). Draw the figure described above and change the color of
the center of the incircle to the smearing color.
Then move point A and see what happens.

The result of one such experiment is shown

in figure 6.1. The incenter sweeps out two
curves depending upon which side of the chord
BC it lies. The two curves look like circles,
Figure 6.1: Locus of Incenters
but if you would like more evidence, try creatLocus/InLocus.T [M]
ing three points on the smeared curves and see
if the circle through them appears to match the
rest of the points. (See Section 6.1.1 for the details on how to do this using Geometer.)
You can even find the center and radius of that circle, and perhaps that extra evidence
will help you to prove your result.
It is also a good idea in situations like this to repeat the experiment with different
configurations of the points B and C to see how that affects the locus curve.

6.1.1 Technical Details

If you would like to place a circle on the smeared images of the point that traces the
incenter of 4ABC in figure 6.1 using Geometer, you will find that the instant you create a new geometric object, all the smear points disappear (because Geometer thinks
you are smearing out a new set). Here is a reasonably easy approach to the problem.
Move A to smear out the curve. Then use the Free P command to create a free
point. Place the cursor over the smear curve and click down. This creates a point, but
the instant the point is created, the smear data disappears. Just smear it again (moving
point A onlyleave B and C where they are), and then create another new point in
the same way. Repeat a third time and you will have three points roughly along the
apparent curve. Draw a circle through those three points and smear a final time by
moving point A. If the smear seems to stick to your circle, it is likely that the desired
locus is, indeed, a circle.
It works better if the three points are as far apart as possible on the smear curve, and
if you have clearly missed the curve with one of the points, either drag it closer to the
smear curve or erase and re-draw it before you create the circle. In fact, it is obvious
that as the point A approaches B or C, so does the center of the inscribed circle, so
you can use points B, C, and a single point on the curve to determine the test circle.
In this example, once you have convinced yourself that the smear curve does seem
to be a circle you can perform other tests on your Geometer circle such as locating its
center, looking for tangent lines, et cetera.

6.1.2 Conjecture and Proof

6.2. EULERS FORMULA

123

When you do draw a circle and see that it seems

to fit the locus of points very closely, you will
find that the center of that circle seems to lie on
the arc BC and in fact seems to bisect the arc.
It
is also obvious that as A gets very close to
I
B (or C), the incenter will also approach B (or
C), so it seems that a good conjecture might be
B
C
that the incenter I of 4ABC satisfies IX =
X
BX = CX. See figure 6.2.
If we test the diagram with Geometer, our
Figure 6.2: Locus of Incenters
suspicions are confirmed: We are told that
Locus/InLocus1.T [P]
IX = BX and in addition that a bunch of sets
of angles are equal. Most of the angle equalities
are obvious from the figure, but Geometer also reports that XIB = XBI.
Using these ideas, it is simple to prove this lemma. AX is an angle bisector so
BAX = CAX. In addition, AX will bisect the arc BC, so X, the proposed
center of our circle of locus points, is in the center of that arc. Of course CBX is
also inscribed in the arc CX so it is also equal to BAX and CAX.
The line BI is also an angle bisector, so CBI = IBA. We also know that
XIB is equal to IAB + IBA since it is an exterior angle of 4AIB. But clearly
XBI = XBC + CBI, and since both are equal to the sum of congruent angles,
XIB = XBI. Thus 4XIB is isosceles, and hence IX = BX.

6.2 Eulers Formula

E
A

T
I O

Q
C

B
X

The locus sought in Section 6.1 perhaps seems

a bit artificial, but in fact it might be one of the
first things you would try if you were working
on the following problem:
If R and r are the radii of the inscribed and
circumscribed circles of a triangle, show that
the distance d between the centers of those circles satisfies the equation:
R2 d2 = 2Rr.

(6.1)

This is known as Eulers formula1, and is

closely related to the triangular version of PonLocus/InLocus2.T [P]
celets theorem. See Section 3.8.
A reasonable approach is to leave R constant and check a variety of inscribed triangles to see how r and d might related. Our experiments with figure 6.1 in the previous
section yield some clues.
Figure 6.3: Locus of Incenters

1 There are probably a dozen formulas called Eulers formula so which one is meant generally depends
upon the context.

124

CHAPTER 6. LOCUS OF POINTS

We also notice that R2 d2 = (R +d)(R

A
d) on the left side of equation 6.1. In figure 6.3
B
d = IO, so it is clear that IQ = R + d and
A B
P I = R d. The point I divides a diameter
into pieces whose product we are seeking. Any
D C
other chord through I will do the same thing,
so we can conclude that R2 d2 = AI IX,
C
D
for example.
On the right side of equation 6.1 we have
the product 2Rr. The diameter of the circum- Figure 6.4: Rectangle of Incenters
circle is 2R, and r is the radius of the incircle.
Locus/Circlecenters.T [P]
If these numbers are going to appear in our calculations, we will need to have segments of those lengths in the diagram. We draw
diameter EX and radius IT as in figure 6.3. Other radii and diameters would also
work, but EX involves the points O and X that are already in the diagram, and IT
connects to the segments AI and IX which appear in formulas we already have. If in
addition we draw EC and XC it appears that 4ECX is similar to 4AT I. Of course
we can use Geometer along the way to test for the various similarities.
With these constructions, the proof of Eulers formula is easy.
Clearly AT I = 90 since IT is
A
a radius of the circle to which AT is
tangent, and ECX = 90 since it
B
is inscribed on the diameter EX. By
A
AA, 4ECX 4AT I so AI/IT =
C
C B
EX/CX. But the lemma proved in
G
B
Section 6.1.2 tells us that CX = IX,
A
IT = r, and EX = 2R, so if we
C
simply substitute and cross-multiply,
A B C
we find that 2Rr = AI IX = (R +
d)(R d) = R2 d2 .
If you are interested in using the
Figure 6.5: Three Medians
lemma proved in Section 6.1.2 again,
Locus/Medians3.T [P]
look at the diagram in figure 6.4.
Show that the four incenters of the triangles obtained by adding a diagonal to a cyclic
quadrilateral lie at the corners of a rectangle. We will not prove it here in the text, but
the Geometer diagram contains the complete proof.

6.3 Triangle from Three Medians

Given the lengths of the three medians of a triangle, construct the triangle. This seems
like a difficult problem at firstnone of the sides or angles of the original triangle are
given, so it is hard even to know how to begin.
We do know a few facts about medians. All three medians of a triangle meet at
the centroid, G, and G is 2/3 of the distance from each vertex to the opposite side.

6.3. TRIANGLE FROM THREE MEDIANS

125

In addition, since they are medians, the end of each median meets the midpoint of the
opposite side.
The next thing to do is to draw a Geometer program to get a feel for the problem.
Draw your own, or load the file Locus/Medians3.T. The file used here has three lengths,
AA0 , BB 0 and CC 0 that can be adjusted on the left of the main diagram. The centroid,
G, is fixed, and all three copies of the lengths on the left are copied and constrained
to pass through G at 2/3 of the distance from the vertices-to-be (A, B and C). Since
any one of them can be fixed, in the diagram CC 0 is not movable. The points A and B
can be moved, each on a circle centered at G with radius equal to 2/3 of the lengths of
AA0 and BB 0 , respectively. Figure 6.5 illustrates a situation where A and B have been
manually adjusted to be almost correctthe points A0 , B 0 and C 0 lie close to, but not
exactly on, the edges BC, CA and AB.
You can get a feeling for the problem and even obtain approximate solutions by
experimenting with a Geometer diagram, but this will not provide a construction. So
far the diagram has not made use of the fact that the endpoint of each median has to lie
on the midpoint of the opposite side. We are leaving C fixed, but what if we allow B to
move completely around its circle centered at G and having radius BG? Every location
of B on that circle of allowable positions determines where B 0 lies. The locations of
B 0 and C determine where A would have to be in order that the location of B be a
solution. Since B 0 is the midpoint between A and C in the solution, we can reflect C
across B 0 to find where A would have to lie if B (and hence B 0 ) were in the correct
position. That theoretical position of A is shown as A in figure 6.6 for one particular
configuration of B, G and C.

A*

A
B

A
C

A B C

B C
G
A
C

In addition, A must lie on the circle centered at G and having radius

(2/3)AA0 . If A is drawn in a smearing color and B is moved around its
circle we see the locus of possible locations of A, which appears to be a
circle. Since the actual position of A
has to be on that circle and on the circle of radius (2/3)AA; about G, the
solution puts A at the intersection of
the two circles.

A quick analysis shows us that the

point A is the reflection of B through
the point C 0 . Thus if B sweeps out a
Locus/Medians3.T [P]
circle of radius r, then A will sweep
out a circle of the same radius. A
brute-force method to construct this circle would be to select three different locations
for B on its circle around G and reflect each of them through C 0 . Then construct the
circle passing through those three points. The intersection of that circle and the circle
of radius (2/3)AA0 about G is the location of vertex A. From there it is easy to find B
and the construction is complete.
Figure 6.6: Three Medians

126

CHAPTER 6. LOCUS OF POINTS

To review the construction: Draw the segment CC 0 at any location. Find the centroid G that is 2/3 of the distance from C to C 0 . Draw a circle about G of radius
(2/3)BB 0 and select three points on it. Reflect these three points through C 0 and
construct a circle passing through the three reflected points. Draw a circle of radius
(2/3)AA0 about G and find an intersection of it with the circle through the three reflected points. This is the location of A. Reflect A through C 0 to obtain the location of
B. This completes the construction.

6.4 The Ellipse

An ellipse is commonly defined to be the set of points the sum of whose distances to
two fixed points (the foci) is constant. Although Geometer draws ellipses (and in fact,
all conic sections) as primitive objects, it is easy to use the definition above to construct
as many points as desired on an ellipse using only line and circle drawing commands.
A diagram to draw an ellipse as in
figure 6.7 is quite simple. The two
a
L-a
foci, F1 and F2 are completely free
points. Next, a line segment is drawn
that has total length L, where L repF1
F2
resents the sum of the distances to the
foci. A point constrained to be on the
line segment divides it into two segments of length a and L a. A circle
of radius a is drawn about F1 and a
Figure 6.7: Ellipse from Definition
circle of radius L a is drawn about
Locus/Ellipse.T [P]
F2 . The points at the intersections of
the two circles are drawn in a smearing color. As the point separating the two segments is moved, the relative sizes of the
circles changes and so do their intersections which smear out the ellipse. Such a diagram makes it easy to adjust the distances between the foci and total length L to see
how those changes affect the resulting ellipse.
The ellipse in figure 6.7 is not very pretty since at the left and right ends when the
circles approach tangency there are not nearly as many intersections. Remember that
the point on the line can only move in units of one pixel on the computer screen so the
radii of the circles also change in one-pixel increments. When the circles are close to
tangency, making one circle larger and the other smaller by an entire pixel will cause
the points of intersection to jump significantly.
Perhaps the easiest way to improve the drawing is to recall that an ellipse is simply
a scaled circle. In other words, if every point on a circle has its x-coordinates multiplied
by one value and its y-coordinates by another, the resulting points will lie on an ellipse.
In figure 6.8 we have done just that. The point A is constrained to be on a circle, and
with the editor we added a single line to the end of the file. Here is the Geometer code
in its entirety, and only the final line was created with the text editor.

6.5. LOCUS CONSTRUCTION EXERCISES

127

.geometry "version 0.60";

v1 = .pinned(0, 0);
v2 = .free(0.95, 0, .in);
c1 = .c.vv(v1, v2);
v3 = .vonc(c1, -0.850982, 0.422292, "A");
ve = .v.vscale(v3, 1.700000, 0.400000, .smear, .dot);

The final line tells Geometer to

determine the coordinates of ve by
multiplying the x-coordinate by 1.7
and the y-coordinate by 0.4.

The easiest way to get even spacing is simply to use a parametric definition of the ellipse. If you want an
ellipse centered at the origin, aligned
with the coordinate axes, and passing through (, 0), (, 0), (0, )
Figure 6.8: Ellipse with Scaling
and (0, ), then as the parameter
Locus/Ellipse1.T [P]
t varies from 0 to 360 , the points
( cos t, sin t) will sweep out the ellipse. If t takes uniform steps, the output points will not be spaced uniformly, but they
will be spaced in a pleasing, symmetric manner.
Using the Geometer .script command below causes the value of t to go from 0
to 360 in steps of 2. When this script is run by pressing the Run Script button, the
drawing in figure 6.9 is generated.
.geometry "version 0.60";
t = .script(0.000000, 360.000000, 2.000000);
x = .f.rpn(t, .cos, 1.700000, .mul);
y = .f.rpn(t, .sin, 0.400000, .mul);
v = .v.ff(x, y, .smear, .dot);

6.5 Locus Construction Exercises

Without any explanation of the results, here are a few problems where
the goal is to find a locus of points.
Read the problems, try to figure out
the locus (with or without proof), and
then check the associated Geometer
Figure 6.9: Ellipse with Script
files for the authors solution. ReLocus/Ellipse2.T [P]
member that using the Edit Geometry
command you can look at the exact order that the commands were issued in Geometer.
All the examples below were constructed using only Geometers graphical user
interface; it was never necessary to edit the textual version of the diagram.

CHAPTER 6. LOCUS OF POINTS

128

1. Let A, B, C and D be any four points on a circle. Let P be any point (inside,
outside, or on the circle), and draw the lines P C and P D. Let Q and R be the
intersections of P C and P D with the circle other than C and D. Find the locus
of the intersections of the circles P RA and P QB. (Solution: Locus/Locus1.T)
2. Let B and C be two fixed points on a circle, and let A move freely on the circle. Find the locus of the projections of the orthocenter H onto the bisector of
CAB. Depending on what you mean by the angle bisector (there is an interior
and exterior bisector) this could mean different things. The included Geometer
diagram changes the interpretation from interior to exterior when the point A
changes sides of the segment. You may wish to draw both the interior and exterior bisectors to see what happens in both cases. (Solution: Locus/Locus2.T)
3. Given a fixed circle K, find the locus of the endpoints of the diameters parallel to
a given line of a fixed-radius circle orthogonal to K. (Solution: Locus/Locus3.T)
4. Given a fixed circle K and a fixed point P , allow Q to move around K, and
construct the circle through P that is tangent to K at Q. What is the locus of the
intersection of the lines tangent to that new circle that pass through P and Q?
(Solution: Locus/Locus4.T)
5. Given two points A and B, find the locus of points P such that P B : P A = r : 1,
where r is a fixed ratio. (Solution: Locus/Locus5.T)

6.6 Constant Ratio

The final exercise in the previous section is important enough to deserve a proof. After
constructing the locus using Geometer, it is clear that the answer is a circle, but the
proof is a little tricky.
C

Figure 6.10: Interior and Exterior Subdivision

Locus/InEx.T [P]

Before stating the theorem we will first prove that the internal and external bisectors
of an angle of a triangle divide the side opposite internally and externally in the same
ratio. In figure 6.10, if CI amd CI 0 are the internal and external bisectors of C, then
AI/BI = AI 0 /BI 0 . This is just a taste of what is to come in Chapter 10.

6.7. HIGHER-ORDER CURVES

129

The proof is not difficult. Let ACI = BCI = . Since CI is the angle bisector
of ACB, we have AI/AC = BI/BC. Notice also that since CI and CI 0 are both
bisectors, ICI 0 = 90 . Next, we apply the law of sines to triangles 4ACI 0 and
4BCI 0 :
AI 0
AC
=
sin(90 + )
sin CI 0 B

and

BI 0
BC
=
.
sin(90 )
sin CI 0 B

Since sin(90 +) = sin(90 ) = cos(), the two equations above can be

combined to yield the desired result: AI/BI = AI 0 /BI 0 .
Now we are ready to prove the main theorem.
Theorem 6.1 (Constant Ratio Theorem) Let A and B be two points on a line and let
I lie on the segment between them such that I is not the midpoint of AB. Then the
locus of points P such that P A/P B = IA/IB is a circle passing through I whose
center is on the line AB. If O is the center of that circle, then2 OA : OI = OI : OB.
Based on the previous result, the proof
is not difficult: see figure 6.11.

Once the locations of the points A, B

and I are chosen, find an arbitrary point P
such that P A/AI = P B/BI. Since this
A
I
B
O
I
ratio holds, that means that P I is the angle bisector of AP B. Construct a line
perpendicular to P I through P passing
Figure 6.11: Points of Constant Ratio
through the line AB at I 0 . P I and P I 0
Locus/ConstRatio.T [M]
are the internal and external angle bisectors of AP B in 4AP B, so the points I and I 0 divide the segment AB internally and
externally in the same ratio. Notice that this makes the position of I 0 independent of P ,
so for every such P , the angle IP I 0 = 90 . Thus P lies on a circle whose diameter
is II 0 , and the proof is complete.

6.7 Higher-Order Curves

In Euclidean geometry, we normally deal with straight lines and circles, and sometimes
conic sections: ellipses, hyperbolas and parabolas. All of these are first degree (lines)
or second degree (all the rest). By first degree we mean that they can be described
by equations whose highest power is 1. Second degree curves have equations whose
highest power is 2, et cetera. The general equation for any straight line is given by:
ax + by + c = 0,
2 We will see in Chapter 8 that A and B are inverses of each other relative to the circle. Also note that
if I is at the midpoint of the segment AB, then the locus is the perpendicular bisector of AB which can be
thought of as a circle of infinite radius.

CHAPTER 6. LOCUS OF POINTS

130

for suitably-selected constants a, b and c. The general second degree equation is:
ax2 + bxy + cy 2 + dx + ey + f = 0,
where a, b, . . . , f are constants. Second degree equations describe all the conic sections, where we consider a circle to be a special form of an ellipse.
We have already encountered two examples of curves of higher degree than 2 (the
cardioid and the limacon) in Section 2.10.2.

6.7.1 Example: Strophoids and Freeths Nephroid

Given a curve C and two points, A and O, let X
be an arbitrary point on C, and let P1 and P2 be
the two points on the line XO such that AX =
AP1 = AP2 . The locus of all such Pi is called
the strophoid of the curve relative to O and A.

A
O

In figure 6.12, the curve C is a circle, O is

the center of the circle, and A lies on the circle.
This particular strophoid is called the nephroid of
Freeth.
Figure 6.12: Freeths Nephroid
Try drawing strophoids of other curves (do
Locus/Strophoid.T [S]
not forget the straight line), and modify Freeths
nephroid to a more general strophoid by allowing the point A to be off the circle, and/or
make O different from the center of the circle.

6.7.2 Additional Higher-Order Examples

Following is a collection of interesting looking curves. To save space, the drawings are
not included here, but every one is available as a Geometer diagram on the CD where
each curve is drawn by pressing the Run Script button. In every case it is probably
worthwhile to try to draw the curve yourself using Geometer before loading the file to
see the solution.
 In many cases in the examples that follow, multiple descriptions may be given for
the curve, including some combination of geometric, cartesian, polar, and parametric.
If you do not know what these terms mean, do not worryyou can still draw the
pictures by pressing the Run Script button.
Many additional examples of interesting curves can be found in [Lawrence, 1972].
6.7.2.1 The Cissoid of Diocles
Given a line L and a circle C tangent to it at Q. Let S be the point at the other end
of the diameter of C through Q. If X is a point on L and R is the intersecion of the
line XS with C, then the cissoid of Diocles is the locus of points P on XS such that
P X = RS.

6.7. HIGHER-ORDER CURVES

131

Cartesian equation: y 2 (a x) = x3 .
Polar equation: r = a sin tan .
Parametric equation: (a sin2 t, a tan t sin2 t).
Geometer file: Locus/Cissoid1.T
6.7.2.2 The Parabola
Given a line L and a point F (the focus) not on L, a parabola is the set of all points
whose distance from the focus F and the line L are equal.
Cartesian equation: y 2 = 4ax.
Polar equation: r sin2 = 4a cos .
Parametric equation: (at2 , 2at).
Geometer file: Locus/Parabola.T
6.7.2.3 The Conchoid of Nicomedes
Given a line L and a point O, let the point Q move along L. The conchoid of
Nicomedes is the locus of points P1 and P2 on the line OQ such that P1 Q = P2 Q.
Cartesian equation: (x2 + y 2 )(x b)2 = a2 x2 .
Polar equation: r = a + b sec t.
Parametric equation: (b + a cos t, tan t(b + a cos t)).
Geometer file: Locus/Conchoid.T
6.7.2.4 The Hypotrochoid and Rhodonea
If a small circle rolls around the inside of a larger circle and the ratio of their radii is an
integer (so that the small circle completes an integral number of turns when it returns to
its starting point), then if P is a point connected to the smaller circle, a hypotrochoid
is the path traversed by P . Note that P may be outside the smaller circle, but if that is
the case, it rotates as if it were connected to the smaller circle by a rigid rod.
A rhodonea is a special case of the hypotrochoid where if h is the distance of
P from the center of the smaller circle, the radii of the smaller and larger circles are
h(n 1)/(n + 1) and 2nh/(n + 1), respectively.
Polar equation: r = a cos(n).
Geometer file: Locus/Rhodonea.T (for n = 7)
6.7.2.5 Lissijous Figures
If two wheels of the same size turn at different rates such that they eventually return to
the starting position after a certain number of turns, and if a point is selected on each,
then a lissijous figure is the locus of points that are the intersection of a vertical line
drawn from one of the points and a horizontal line from the other.
Parametric equation: (cos nt, sin mt), (cos nt, cos mt) or (sin nt, sin mt).
Geometer diagram: Locus/Lissijous.T

CHAPTER 6. LOCUS OF POINTS

132
6.7.2.6 Dumbbell Curve
This is simply a curve that looks like a dumbbell.
Parametric equation: y 2 = x4 x6 .
Geometer file: Locus/Dumbbell.T

6.7.2.7 Bezier Splines

Choose a set of n 3 control points. Given t, where 0 t 1, between each pair,
choose a point that divides the segment connecting them into a ratio of t and 1 t,
yielding n 1 points. Repeat the process until there is only a single point. As t varies
from 0 to 1, this final point sweeps out a curve called a Bezier spline. Geometer
provides cubic Bezier splines as a drawing primitive.




Pn
Pn
Parametric equation: ( i=0 xi ni ti (1 t)ni , 0 yi ni ti (1 t)ni ), where the
coordinates of the control points are (xi , yi ).
Geometer file: Locus/Bezier.T (with 5 control points)

6.8  Plotting Curves of the Form: y = f (x)

By default, Geometers coordinate system runs from 1 to 1 in the windows smaller
dimension, so (0, 0) is always in the center of the window. Suppose we would like to
prepare a nice drawing of the function:
y=

cos(x)sin(10x)
+ sin(3x)
2

from x = to x = , where the angles are measured in radians.

Figure 6.13: y = cos(x)sin(10x)/2 + sin(3x)

Locus/Plot.T [S]

The final graph will occupy a space 2 6.28 units long and as much as 3 units
high. The smaller dimension will be the 3 units, and it has to fit in a coordinate system
running from 1 at the bottom to 1 at the top. Thus we will scale everything by a factor
of 2/3.
Here is the Geometer code:

6.9. PLOTTING CURVES IN POLAR COORDINATES

133

.geometry "version 0.60";

.radianmode;
t = .script(-3.141590, 3.141590, 0.010000);
x = .f.rpn(t, 0.666667, .mul);
y = .f.rpn(t, .cos, t, 10.000000, .mul,
.sin, .mul, 2.000000, .div, t,
3.000000, .mul, .sin, .add, 0.666666,
.mul);
v = .v.ff(x, y, .smear, .dot);

The .radianmode makes Geometer calculate in radians instead of degrees. The t

varies from to . The true x value is t, but to fit the curve to the Geometer window
we need to multiply it and the y-coordinate by 2/3. The y-coordinate is calculated in
a straightforward way, and the result is plotted in a smearing color. The window size
is adjusted to have a ratio of about 2 long to 3 high, and the Run Script button yields
the curve shown in figure 6.13.

6.9  Plotting Curves in Polar Coordinates

If the points on a curve are defined as a relationship between the distance r from the
origin and the angle the ray from the origin to the point makes with the positive xaxis, the curve is said to be described in polar coordinates. For example, a circle of
radius 1 centered at the origin has a radius independent of the angle, so the polar form
of that circle is r = 1. Usually things are not that simple, however.
There are two ways to plot such curves using Geometer. One is for each (r, ) pair
to draw a circle of radius r and a ray at angle and plot (with a point having the smear
color) the intersection of the ray and circle. If you do not mind editing the textual
version of a Geometer file, another method is to calculate the cartesian coordinates
(x, y) = (r cos , r sin ) and to plot those points as r and vary appropriately.

Figure 6.14: Archimedes Spiral

Locus/Polar.T [S]

Suppose we wish to plot Archimedes spiral given by the polar equation r =

(0.001), where the angle is measured in degrees. Two solutions to the problem
are presented below in Geometer code format. The two files are Locus/Polar.T and
Locus/Polar1.T and are listed below. The listing of Locus/Polar.T is on the left. Both
are scripts, so run them by pressing the Run Script button. See figure 6.14.

CHAPTER 6. LOCUS OF POINTS

134
.geometry "version 0.60";
o = .pinned(0, 0, "O");
one = .pinned(1, 0, .in);
l1 = .l.vv(o, one, .longline);
l2 = .l.vlperp(o, l1, .longline);
theta = .script(0.0, 1080.0, 4.0);
r = .f.rpn(theta, 0.001, .mul);
x = .f.rpn(r, theta, .cos, .mul);
y = .f.rpn(r, theta, .sin, .mul);
p = .v.ff(x, y, .smear, .dot);

.geometry "version 0.60";

o = .pinned(0, 0, "O");
one = .pinned(1, 0, .in);
l1 = .l.vv(o, one, .longline);
l2 = .l.vlperp(o, l1, .longline);
theta = .script(0.0, 1080.0, 4.0);
r = .f.rpn(theta, 0.001, .mul);
c = .c.vf(o, r);
ang = .a.f(theta);
v1 = .v.avv(ang, one, o);
l3 = .l.vv(o, v1);
p = .v.lc(l3, c, 2, .smear, .dot);

6.10  Envelopes of Curves: the Epicycloid

Sometimes the shape of a curve is made obvious if you draw a large set of lines that are
tangent to the curve. The various types of epicycloids provide some good examples.
See figure 6.15.
An epicycloid is path traced by a point on the edge of a circle as it rolls around
another circle. If the ratio of the diameters of the circles is rational, the path closes.
To draw the envelope of an N -epicycloid begin by selecting a reference point on
a circle. Next, for an evenly-spaced set of points around the circle, connect each of
those points by a straight line to the point that is N times as far around the circle as
the original point, wrapping around the circle if necessary. Figure 6.15 illustrates the
results for N = 2, N = 3, N = 5 and N = 8.
The three examples were generated by the following Geometer code:
.geometry "version 0.61";
m = .script(0.000000, 360.000000, 2.000000);
rad = .f.rpn(0.800000);
vn2 = .pinned(-0.9, 0.92, .dot, "2");
vn3 = .pinned(-0.72, 0.92, .dot, "3");
vn4 = .pinned(-0.54, 0.92, .dot, "4");
vn5 = .pinned(-0.36, 0.92, .dot, "5");
vn6 = .pinned(-0.18, 0.92, .dot, "6");
vn7 = .pinned(0, 0.92, .dot, "7");
vn8 = .pinned(0.18, 0.92, .dot, "8");
vn9 = .pinned(0.36, 0.92, .dot, "9");
vn10 = .pinned(0.54, 0.92, .dot, "10");
vn11 = .pinned(0.72, 0.92, .dot, "11");
vn12 = .pinned(0.9, 0.92, .dot, "12");
l1 = .l.vv(vn2, vn12);
vmid = .vonl(l1, -0.365269, 0.92, .red, "N");
v2x = .f.vxcoord(vn2);
v12x = .f.vxcoord(vn12);
vmidx = .f.vxcoord(vmid);
n = .f.rpn(vmidx, v2x, .sub, v12x, v2x,
.sub, .div, 10.000000, .mul, 0.500000,
.add, .truncate, 2.000000, .add);
p0 = .pinned(0, 0, .in);
p1 = .v.ff(rad, 0.000000, .in);
c = .c.vv(p0, p1);

6.10. ENVELOPES OF CURVES: THE EPICYCLOID

135

2N

10

11

12

3N

10

11

12

5N

10

11

12

8N

10

11

12

Figure 6.15: Various Epicycloid Envelopes

Locus/Epicycloid.T [S]
vx0 = .f.rpn(m, .cos, rad, .mul);
vy0 = .f.rpn(m, .sin, rad, .mul);
v0 = .v.ff(vx0, vy0, .in);
m1 = .f.rpn(m, n, .mul);
vx1 = .f.rpn(m1, .cos, rad, .mul);
vy1 = .f.rpn(m1, .sin, rad, .mul);
v1 = .v.ff(vx1, vy1, .in);
l = .l.vv(v0, v1, .smear);
.text("Epicycloid: Set N above and press Run Script.");

Here is how the code works. The variable m is the angle around the circle of the
initial point, measured in degrees. It varies from 0 to 360 in steps of 2 . This can be
changed to a smaller number for more tangent lines. It can be any value, but it makes
a more uniform diagram if it divides evenly into 360. The value of rad is the radius of
the circle which is centered at the origin (center = p0; point on boundary = p1).

136

CHAPTER 6. LOCUS OF POINTS

The next few lines of code allow the user to choose the value of N for which the
N -epicycloid will be generated. The point vmid (labeled N in the Geometer diagram)
can be moved between vn2 and vn12 and its position is used to calculate n.
To calculate n we find the x-coordinates of vn2, vmid and vn12. The expression
(vmidx - v2x)/(v12x - v2x) is the proportion of the distance along the line from vn2 to
vn12 at which vmid lies. It will be between 0.0 and 10.0. When we multiply by ten,
add 0.5 (for rounding) and truncate, we obtain an integer between 0 and 10. Adding 2
to that gives us an integer n between 2 and 12, inclusive.
The points labeled vn3, . . . , vn11 are simply to label the line so the user can tell
what value of n will be calculated.
Next the circle is drawn and the coordinates of the starting point (determined by
angle m) and the endpoint (determined by angle mn) of the line segment are calculated.
That line is then drawn in a smearing color.

Chapter

Triangle Centers
Perhaps the most amazing theorems from high school geometry concern the concurrence of the medians, altitudes, angle bisectors and even the perpendicular bisectors
of the sides of a triangle. This occurs not just for particular triangles, but for every
triangle. These high school examples all have special names that may at first sight
seem unrelated to how they are defined. The intersection of the three medians is called
the centroid, the intersection of the three angle bisectors is called the incenter, the
intersection of the three altitudes is the orthocenter, and the intersection of the three
perpendicular bisectors of the sides is the circumcenter.
As we shall see, the points of concurrence above just scratch the surface. These, and
all the similar points that will be defined later in this chapter will go by the general name
triangle center. We will learn about more of these centers, about other properties that
the centers satisfy, interrelationships among the centers, and how the various centers
have applications not only to other areas of geometry, but very practical applications as
well.
137

138

CHAPTER 7. TRIANGLE CENTERS

A computer geometry program provides an excellent tool to study these triangle

centers because after the geometric definition is made, you can use the computer to investigate how the centers are related to the size and shape of a huge variety of triangles.
Without a computer it would be important to draw many careful examples by hand to
learn about these relationships.
In fact, if you use Geometer, there is a program on the CD called Finder.T that
provides a tool to help determine the identity of an unknown center. Just construct
your center in the diagram and then page through a large collection of common centers
(including all those covered in this chapter) to see which one seems to be the same as
yours. See Section 7.12 for details.
To prove that various points are centers of one form or another we will often need
to show that three lines are concurrent. One of the most useful and powerful ways
to do this is with Cevas theorem, so now might be a good time to review that. See
Section 5.1

7.1 General Properties of Triangle Centers

We will be looking at a lot of specific triangle centers in what follows, but since we
have already mentioned a few of them, it is a good idea to take a look at the sorts of
properties that all triangle centers have. Here is a short list of those properties. You
know about the four of them mentioned above (the incenter, circumcenter, orthocenter,
and centroid)check to see that these properties hold at least for those four centers.
The center will exist for triangles of every shape. The term center may be a
little misleading in that there is no guarantee that it even lies inside the triangle
the orthocenter and the circumcenter, for example, may lie outside.
Similar triangles will have similar centers. In other words, if you rotate or translate a triangle, or uniformly scale it to a larger or smaller size, the center should
rotate or translate with the triangle, or should scale up and down as the triangle
does.
As a sort of corollary to the statement above, a triangle center must depend
equally on the three vertices. In other words, it would not be reasonable to
define a new kind of center for 4ABC to be the midpoint of the segment AB
because a simple relabeling of the vertices of the triangle would move this supposed center to lie on a different side of the triangle.

7.2 The Gergonne Point

7.2. THE GERGONNE POINT

139

B
Z
A

G
Y

If the points of tangency of the incircle of

triangle 4ABC are connected to the opposite vertices as in figure 7.1, then they
all meet at a point, called the Gergonne
point.

X
C

Figure 7.1: The Gergonne Point

Tricenters/Gergonne.T [M]

7.2.1 Geometer Analysis of the Gergonne Point

This is not too difficult to prove, but with Geometers ability to search for relationships,
it is extremely easy. When we test the diagram for the figure above, the following
results are returned:
Points on a line:
C Z G
Concurrent lines:
(AX) (BY) (CZ)
Equal length segments:
[C X] [C Y]
Equal length segments:
[B Z] [B X]
Equal length segments:
[A Z] [A Y]
Ratios:
[B Z] x [C X] = [B X] x [C Y]
[A Z] x [C X] = [A Y] x [C Y]
[A Z] x [B Z] = [B X] x [A Y]

Geometer agrees that the lines AX, BY and CZ are concurrent, but it also reports
some interesting information about ratios and equal lengths.
We recall that perhaps the most common method to show that three lines are concurrent is to apply Cevas theorem. In the problem above, the three lines will be concurrent
if
AZ BX CY

= 1.
BZ CX AY
We notice that in each pair of equal segments reported by Geometer, one appears in
the numerator above and one in the denominator, and that each segment length appears
exactly once. Thus if we can show that the three segment length equalities hold, we
will be done. But these are trivial to show, since we know that the lengths of the two
external tangents to a circle from a point (like AY and AZ, for example) are of equal
length.

140

CHAPTER 7. TRIANGLE CENTERS

7.3 The Nine Point Center

Let 4ABC be an arbitrary triangle. If A0 ,
B
B 0 and C 0 are the midpoints of the sides,
X, Y and Z are the feet of the altitudes,
and K, L and M are the midpoints of the
L
X
segments HA, HB and HC where H is
C
the orthocenter (the intersection of the three
A
Z
altitudes), then the nine points A0 , B 0 , C 0 ,
H
K
X, Y , Z, K, L and M all lie on a circle
M
A
called the nine point circle. By manipuY
B
lating a diagram, notice that this holds no
C
matter whether the orthocenter is inside or
outside the triangle. Also notice that the order of the points on the circle may change,
Figure 7.2: The Nine Point Circle
and sometimes two of the points may coinTricenters/Ninepoint.T [M]
cide, but no matter what happens, all nine of
them remain locked on a circle. The center of that circle is a triangle center called the
nine point center. See figure 7.2.

7.3.1 Geometer Analysis of the Nine Point Circle

If we test the diagram as it is drawn
B
in figure 7.2, Geometer finds many relationships, but nothing too surprising.
There are the obvious sets of points
that lie on circles like CXY H and
L
X
BXZH, but this is due to the fact that
C
the altitudes make right angles with the
A
bases. Similarly, there are a lot of raZ
H
tios, but most are due to the fact that
K
we have sets of secants cutting through
the circle. We will probably have to
M
A
Y
construct some lines to make progress.
B
If you erase the nine point circle
C
from the figure to minimize the clutter and just move around the vertices
Figure 7.3: The Nine Point Circle
A, B and C of the triangle, you may
Tricenters/Ninepoint1.T [M]
notice that certain sets of four points
seem always to lie on a rectangle, such
as A0 C 0 KM , A0 B 0 KL and B 0 C 0 LM . This is easy to checkjust draw in one of them
in (say C 0 A0 M K) and then test the diagram again.
When you do this, Geometer does report that the line segments appear to be perpendicular, so it probably is a rectangle, but it reports that C 0 A0 , KM , and the base
AC are parallel and also that A0 M , C 0 K and the altitude BY are also parallel.

7.4. THE FERMAT POINT

141

A look at the diagram shows us why: all those segments connect midpoints of
triangle edges, so they must be parallel to the bases of those triangles and equal to half
their length. In this case, the triangles in question are 4ABC, 4AHC, 4BHC and
4BHA. What is more, the altitude BY is perpendicular to the base AC, so instead of
merely being a parallelogram, A0 C 0 KM is a rectangle.
Obviously the same method can be used to show that other sets of four points also
lie on rectangles.
We can find the center of the rectangle if we wish by finding the intersection of the
diagonals A0 K and C 0 M , and in fact it should be clear that both of those segments
are diameters of the circle we are looking for. But then we notice that if A 0 K is the
diameter of a circle and A0 XK = 90 then X must lie on the circle with diameter
A0 K. From here, the proof is obvious.

7.3.2 Proof of the Nine Point Circle

See figure 7.3. Connect A0 C 0 and M K. We know that A0 C 0 k AC since A0 C 0 connects
the midpoints of AB and BC in 4ABC. By exactly the same reasoning, KM k AC
using 4AHC, C 0 K k BY using 4BAH and A0 M k BY using 4BCH. Thus
A0 M k C 0 K and A0 C 0 k KM .

We know BY AC since BY is an altitude, and since C 0 K and A0 M are parallel

to BY and A0 C 0 and CM are parallel to AC, C 0 K 0 and A0 M must be perpendicular
to A0 C 0 and KM . Thus A0 C 0 KM is a rectangle.
Using the other bases and altitudes of 4ABC we can conclude in exactly the same
way that A0 B 0 KL and B 0 C 0 LM are also rectangles.
Consider rectangle A0 C 0 KM . Its center is at the intersection of A0 K and C 0 M
and the circle passing through its four vertices has A0 K and C 0 M as diameters. But
KXA0 = 90 since AX is an altitude, so X must lie on the circle having the diameter
C 0 M . Similarly Z must lie on the circle having diameter A0 K. Thus the points A0 , M ,
K, C 0 , Z and X all lie on a single circle.
By similar reasoning with the other two rectangles we obtain two other sets of six
points lying on circles: A0 B 0 KLXY and B 0 C 0 LM Y Z. Since each of these sets of
points shares three points with each of the others, all the points must lie on a single
circlethe nine point circle.

7.4 The Fermat Point

Erect equilateral triangles on the outside of each segment of 4ABC. The line segments connecting the vertices A, B and C to the tips of the opposite equilateral triangles are concurrent at a point O called the Fermat point of the triangle. See figure 7.4.

7.4.1 Geometer Analysis of the Fermat Point

CHAPTER 7. TRIANGLE CENTERS

142
If we test the diagram in figure 7.4 there are
a few interesting relationships (many of the
relationships found by Geometer have been
omitted from this list):

A
C
B O
B

Concurrent lines:
(AA) (BB) (CC)
Points on a circle:
B C A O
Points on a circle:
A C B O
Points on a circle:
A B C O
Equal angles:
(B C B) (A C A)

Figure 7.4: The Fermat Point

Tricenters/Fermat.T [M]

If we can show that the sets of points lie on a circle, we can conclude that the
angles AOB, BOC and COA are supplementary to 60 angles, so each of them must
be 120 . This will show that the three lines meet at a point.
What we need to show, then, is that the sets of points lie on circles. If we could
show, for example, OA0 C = OBC then there would be a circle passing through
OBCA0 .
The Geometer analysis also tells us that BCB 0 = ACA0 , and those angles
might be used to show the similarity of 4ACA0 and 4B 0 CB. These facts together
provide plenty of clues to construct the proof.

7.4.2 Proof of the Existence of the Fermat Point

A
C
BO
C

B
A

Figure 7.5: The Fermat Point

Tricenters/Fermat1.T [M]

Construct the circumcircles of each of the

three equilateral triangles and let O be the
intersection of AA0 and BB 0 . From the figure it appears that O is also on CC 0 but this
is what we need to prove to show that the
three lines are concurrent at O and that O is
therefore the Fermat point.
In figure 7.5, BCA is equal to itself,
and since both angles A0 CB and ACB 0
are equal to 60 , we have:
A0 CA

= 60 + BCA
= BCA + 60 = BCB 0 .

Since the outer triangles are equilateral, CB 0 = CA and A0 C = BC, so from SAS,
we know that 4A0 CA
= 4BCB 0 . Because the triangles are congruent, CB 0 O =
CAO and OA0 C = OBC, so OBA0 C and OACB 0 form two sets of four concyclic points. Therefore O lies on the intersections of the circumcircles of the equilateral triangles 4A0 BC and 4AB 0 C.

7.5. THE NAGEL POINT

143

The same argument can be used to show the congruence of two other pairs of triangles: 4ABA0
= 4C 0 BC and 4B 0 AB
= CAC 0 and their common intersections
also form sets of concyclic points. If we can show that the three circumcircles of the
equilateral triangles pass through a single point we are done since that point must be
O.
Since opposite angles in a cyclic quadrilateral are supplementary and the outer
triangles are all equilateral, AOC = COB = 120 . Thus BOA = 120 since
AOC + COB + BOA = 360. Therefore O lies on the circumcircle of 4ABC 0
since it is opposite a 60 angle and we are done.
There is a second Fermat point that can be obtained if the three equilateral triangles
point inward and not outward. The proof of this is left as an exercise.

7.5 The Nagel Point

A
F
N

Given an arbitrary triangle 4ABC, extend

the sides and construct the three circles that
are exterior to the triangle and tangent to all
three edges (the excircles). Connect each
vertex with the point of tangency of the excircle opposite it. Those three lines all meet
at a single point called the Nagel point.

Figure 7.6: The Nagel Point

Tricenters/Nagel.T [M]

7.5.1 Geometer Analysis of the Nagel Point

Testing the Geometer diagram on figure 7.6 almost instantly shows us how to solve
the problem. The relationships found by Geometer include the following:
Equal length
[B D] [A
Equal length
[B F] [C
Equal length
[A F] [C

segments:
E]
segments:
E]
segments:
D]

One of the most common methods used to prove that three lines are concurrent
is Cevas theorem which states that the necessary condition for the three lines in this
figure to be concurrent is:
AF BD CE

= 1.
BF CD AE
which is a trivial consequence of the relationships discovered by Geometer. The formulas in Section A.6.8 tell us instantly that BD = AE, BF = CE and AF = CD,
so we are done.

CHAPTER 7. TRIANGLE CENTERS

144

7.6 The First Napoleon Point

Given an arbitrary triangle 4ABC, erect equilateral triangles on all three sides pointing outward. If the centers of these equilateral triangles are connected to the opposite
vertices with line segments as in figure 7.7, those three line segments will be concurrent
at a point called the first Napoleon point. (If the equilateral triangles are all drawn
inward from the faces, and the centers connected in the same way, those lines will also
be concurrent in what is called the second Napoleon point.
This is highly reminiscent of the Fermat point (see Section 7.4). The difference
is that the vertices of the original triangle are connected to the outermost vertex of a
different isosceles triangle erected on each side. For the Fermat point, the triangles
were equilateral; in this case, they have angles of 30 , 30 and 120 .

7.6.1 Geometer Analysis of the Napoleon Point

In figure 7.7 it seems reasonable to draw
a few extra lines and to identify some
points since otherwise the bare centers
B
H
of the triangles will be connected to
K
nothing, and since one of the most comG J
C
mon ways to prove the concurrency of
C
L
lines is with Cevas theorem, it would be
I
A
nice to have the points J, K and L idenB
tified so that we may be able to work
with the cevians in 4ABC. Thus we
draw the segments BH, HC, CI, IA,
Figure 7.7: The First Napoleon Point
AG and GB and find points J, K and
Tricenters/Napoleon.T [M]
L that are the intersections of the lines
CG, AH and BI with the sides of the original 4ABC.
If we test this configuration with Geometer, there are quite a few relationships,
most of which are trivially true due to the fact that there are three equilateral triangles, and that when we connect the centers of those triangles to the two vertices that
are shared with 4ABC, we form a large set of similar figures. If we look carefully,
however, we can find a few relationships that are not so obvious:

Equal angles:
(B C I) (A C H)
Equal area triangles:
[B C I][A C H]
Triangle area ratios:
[A B H] x [C H K] = [B K H] x [A C H]

There are equivalent relationships among all the triangles in corresponding positions in the figure, but let us just concentrate for now on this pair.
From the second relation, we can see that the point K divides the base of 4BCH
into two triangles whose areas appear in a ratio. But if we think of BC as the base of

7.6. THE FIRST NAPOLEON POINT

145

4BCH, the ratio BK : CK is equal to the ratio of the areas of the smaller triangles
4BKH and 4CKH. If we are going to use Cevas theorem, we will need to use
ratios involving BK and CK, so perhaps we can convert them to ratios of triangle
areas.
Using the equal angles found by Geometer and the ratios of sides due to the similarity of the equilateral triangles and the included 30 30 120 triangles we can
show prove the equality of the triangle areas found by Geometer.

7.6.2 Proof of the Existence of the First Napoleon Point

Let us begin by showing that A(4BCI) = A(4ACH). Since we have equilateral
triangles erected on the sides of the triangle as well as the 30 30 120 triangles,
it is clear that both BCI and ACH are the sum of ACB and a 30 angle, so they
are equal.
We know that the area of a triangle is equal to half the product of two sides multiplied by the sine of the included angle, so
A(4BCI) =

BC CI sin(BCI)
AC CH sin(ACH)
and A(4ACH) =
.
2
2

Since the two angles above are equal, the ratio of the two areas is therefore:
A(4BCI)
BC CI
=
.
A(4ACH)
AC CH

(7.1)

But since they are both 30 30 120 triangles, we know that 4ACI
4BCH, and thus CI/AC = CH/BC, and this, combined with equation 7.1 tells
us that the ratio of the areas is 1, or in other words, A(4BCI) = A(4ACH).
Exactly the same technique can be applied to show that A(4ABI) = A(4ACG)
and A(4ABH) = A(4BCG).

If we consider BC to be the base of triangles 4BCH and 4BCA, it is clear that

KA and KH divide their areas into equal ratios and the ratio of the areas of 4ACH
and 4ABH will also be the same thing.
From this observation we obtain the following three ratios:
CK
A(4ACH)
=
,
A(4ABH)
BK

A(4ABI)
AL
=
,
A(4BCI)
CL

A(4BCG)
BJ
=
A(4ACG)
AJ

If we multiply the three ratios above together and take into account the equality of
the pairs of triangle areas, we obtain:
CK AL BJ

= 1,
BK CL AJ
and this, together with Cevas theorem, is sufficient to prove that the three lines are
concurrent.

146

CHAPTER 7. TRIANGLE CENTERS

7.6.3 A More General Theorem

Notice in the proof above that we did not make use of the fact that the triangles 4ACI,
4CBH and 4BAG are 30 30 120 triangles, other than the fact that 30 angles
are equal to each other. Thus, if any similar isosceles triangles are erected on the sides
of an arbitrary triangle, the lines connecting the tips of those isosceles triangles with
the vertices of the original triangle are coincident. Exactly the same proof will work,
so this is an alternative proof of the existence of the Fermat point.
Another commonly-seen triangle center of this type is formed by erecting squares
of the sides of a triangle and then connecting the centers of those squares with the
opposite vertex of the triangle. This is exactly equivalent to erecting an isosceles 45
45 90 triangle on each edge and connecting the tip to the opposite vertices.

7.7 The Brocard Points

Given a triangle 4ABC, show that
A
there exists some point such that
AC = CB = BA. (The

character is the Greek letter omega,

and is the standard notation for the Bro
card points.) Find a construction for this
C

point. See figure 7.8.

Every triangle has another Brocard

B
point. The other one (called 0 ) is
obtained by using the opposite angles;
in other words, where : 0 AB =
Figure 7.8: The Brocard Point
0 BC = 0 CA. We will invesTricenters/Brocard3.T [P]
tigate the relationship between the two
points and 0 later in this section.
The solution to this problem is similar in a way to the proof that the three angle
bisectors meet at a point. The proof of that fact was based on the fact that any point
on an angle bisector was equidistant from the two rays of the angle, so the point of
intersection of two angle bisectors must be equidistant from all three sides, and thus it
must lie on the third angle bisector as well.

7.7.1 Geometer Analysis of the Brocard Point

For the Brocard point we can use a similar approach. Suppose we begin by finding
all points such that CB = BA. The set of all solutions will probably lie on
some geometric curve. In other words, any point on that curve will satisfy that angle
equality. It we do a similar analysis to find all points Y satisfying YBA = YAC,
those too will lie on another curve. The point (or points) at the intersection(s) of those
curves will cause all three angles to be equal, and thus to lie on the curve corresponding
to the equality of the third pair of angles.

7.7. THE BROCARD POINTS

T
B

S
C

Figure 7.9: Possible Brocard Points

Tricenters/Brocard2.T [P]

147
Even without any idea how to proceed, we can make a Geometer diagram
of the triangle with a free point (call it
S). Find the angle SCB and copy it to
vertex B forming angle ABT as in figure 7.9. Any point lying on the lines
CS and BT will cause the two angles
CB and BA to be equal. As we
drag the point S around C we notice that
the intersection point seems to sweep
out a circle of its own.

In fact, by adjusting the shape of the

original triangle (see Section 7.7.2) we notice that the circle of possible locations of
passes through B and C and seems to be tangent to the line AB at B.

7.7.2 Technical Details

There are three versions of the Geometer diagram on the CD: Brocard1.T, Brocard2.T
and Brocard3.T. Figure 7.8 was generated with Brocard3.T and in fact contains a
construction of the Brocard point and a Geometer proof that it satisfies the appropriate
angle equalities. Figure 7.9 was generated by Brocard2.T and is a script that sweeps
out a set of evenly-spaced points along the circle of possible solutions.
Brocard1.T contains a figure somewhat like the one you would have constructed
to study the locus of points that cause the two angles to be the same. In this figure
you can move the point S manually and observe how the point moves in response.
Brocard2.T is a modified version of that diagram that moves S automatically, and in
fact the only way to move S in that diagram is with the Run Script button in the control
area.

Finally, if you try to construct the diagram yourself you may find it difficult to

cause the angle at B to have T on the correct side of the ray BA. If it is on the wrong
side, simply reflect the point T across the line AB (using the LP=>P Mirror command
under the PrimitivesPoint pulldown menu). This reflected value of T will cause the
angle to be on the correct side. After reflecting the point, set the original version of the
point T to be the invisible color.
Examine the contents of Brocard1.T and Brocard2.T with a text editor to learn the
details of the construction of those diagrams.

7.7.3 Proof of the Existence of the Brocard Point

Once we learn that the locus of possible locations of the point relative to one of the
triangles sides is a circle, it is easy to see why it is a circle and how to construct it.

CHAPTER 7. TRIANGLE CENTERS

148

C
C
B

Figure 7.9 shows that CB is

inscribed in the arc B, and that
since AB is tangent to the circle
and the segment B cuts off the arc
B, that BA must be equal to
CB. Thus any point on the
circle through C and tangent to AB
at B will cause the angle equality
BA = CB to be satisfied.

It is not difficult to construct the

circle.
We know that its center must
Tricenters/Brocard4.T [M]
lie on the perpendicular bisector of
BC since B and C lie on the circle and we know that since the circle is tangent to AB
at B its center must lie on the line through B perpendicular to AB. Any point lying
on a circle centered at the intersection of those lines and passing through the points B
and C satisfies the angle equality. To find the location of the Brocard point, construct
two such circles relative to different sides and their intersection will be the Brocard
point.
Figure 7.10: Brocard, Median and Symmedian

7.7.4 More on the Brocard Points

As we stated at the beginning of Section 7.7, there are two Brocard points depending
upon which way the angles are measured. Here are some interesting facts about the two
Brocard points and 0 that you may wish to try to prove, possibly using Geometer
to help:
If the lines connecting the vertices of the triangle to form an angle with each
side, and those connecting the vertices of the triangle to 0 form an angle 0 ,
then = 0 .
The line A, when reflected across the angle bisector at BAC, is the line A 0 .
Similarly for the other vertices of the triangle1 .
If O is the circumcenter of 4ABC, then O = O0 and O0 = 2.
The pedal triangle of a point relative to a triangle is the triangle that connects
the projections of the point on each of the sides of the original triangle. The
pedal triangles of and 0 relative to 4ABC are congruent.
The line passing through point A and the Brocard point , the median from
point B and the symmedian from point C all meet at a point. See figure 7.10.
The symmedian is the median reflected across the angle bisector. In the figure,
C 0 is the midpoint of segment AB and C 00 is the reflection of C 0 across the angle
bisector of BCA.
1 Two lines through a vertex of a triangle are called isogonal conjugates if each is a reflection of the
other across the angle bisector. It is an interesting consequence of Cevas theorem that if three Cevians meet
at a point, then so also do their isogonal conjugates.

7.8. THE MITTENPUNKT

149

Ic

A
A

Ib

C C

Ia

Figure 7.12: The Mittenpunkt

Tricenters/Mittenpunkt.T [M]

7.8  The Mittenpunkt

Ic
B
A C
B A
Ib

Ia

Figure 7.11: The Mittenpunkt

In an arbitrary triangle 4ABC, let Ia ,

Ib and Ic be the centers of the three excircles opposite the vertices A, B and
C, respectively. Let A0 , B 0 , and C 0 be
the midpoints of the sides BC, CA and
AB, respectively. Then the lines Ia A0 ,
Ib B 0 and Ic C 0 are concurrent, and meet
at a point called the mittenpunkt.

Tricenters/Mittenpunkt.T [M]

7.8.1 Geometer Analysis of the Mittenpunkt

If we test the diagram in figure 7.11 it does not tell us too much, but there are the
following interesting relationships:
Points
B
Points
A
Points
A

on
C
on
C
on
B

a circle:
I\sub{c} I\sub{b}
a circle:
I\sub{c} I\sub{a}
a circle:
I\sub{a} I\sub{b}

We also realize that our best bet to prove the concurrency of three lines is Cevas
theorem, but to use that, we need a triangle where the lines pass through the vertices
and that is not the case when we begin. We do know that the centers of the excircles

CHAPTER 7. TRIANGLE CENTERS

150

lie on the exterior angle bisectors, so we know that A, B, and C lie on the lines I b Ic ,
Ic Ia and Ia Ib , respectively. A suitable triangle might be 4Ia Ib Ic , so let A00 , B 00 and
C 00 be the points where the lines Ia A0 , Ib B 0 and Ic C 0 intersect Ib Ic , Ic Ia and Ia Ib ,
respectively. See figure 7.12.
The initial Geometer analysis indicated that there are some sets of concyclic points
in the initial diagram, which tells us that there are some angle equalities or at the least,
pairs of angles that are supplementary. However, even after we have added the lines
to make 4Ia Ib Ic as in figure 7.12 and tested it, Geometer does not report any such
relationships among the angles. Why is that?
The reason is that Geometer has no way of knowing that the points A, B and C
lie on the lines of that triangle. When Geometer tests angles, it only tests angles that
consist of pairs of line segments that intersect in known points. If we add the segments
Ia B, Ia C, Ib A, Ib C, Ic A and Ic B to the diagram and test again, suddenly there is
plenty of additional information about equal angles, a small amount of which is listed
below. These equalities are symmetric in that each refers to a different angle in the
triangle 4Ia Ib Ic .
Equal angles:
(A B I\sub{c}) (C B I\sub{a}) (I\sub{c} I\sub{b} I\sub{a})
Equal angles:
(B A I\sub{c}) (C A I\sub{b}) (I\sub{c} I\sub{a} I\sub{b})
Equal angles:
(B C I\sub{a}) (A C I\sub{b}) (I\sub{b} I\sub{c} I\sub{a})

We will also need to make use of the fact that A0 , B 0 and C 0 are the midpoints
of the sides of the original triangle, and to use Cevas theorem, we also have to find
equations that involve the sides Ia B, Ic B, and so on.
A look at the diagram shows us that there are six pairs of triangles that are like
4Ib B 0 C and 4Ib B 00 Ia in that they share an angle, they involve the lengths that are
equal and that will be needed to apply Cevas theorem, and they repeatedly make use
of other lengths and equal angles in the diagram.
If we apply the law of sines to each, Cevas theorem can be applied and the result
proved.

7.8.2 Proof of the Mittenpunkts Existence

Consider the two pairs of triangles (4Ib B 0 C, 4Ib B 00 Ia ) and (4Ib B 0 A, 4Ib B 00 Ic ).
We first note that the lines connecting Ia , Ib and Ic bisect the exterior angles of
4ABC. This, plus a little angle chasing shows us that Ib Ia Ic = Ib AC = and
that Ib Ic Ia = Ib CA = . If we assign the names = CIb B 0 and = AIb B 0
as in the figure, we can apply the law of sines to each of the four triangles mentioned
in the previous paragraph:
sin
sin
=
0
CB
Ib B 0
sin
sin
=
AB 0
Ib B 0

and
and

sin
sin
=
00
Ia B
Ib B 00
sin
sin
=
Ic B 00
Ib B 00

(7.2)
(7.3)

7.9. A NEW CENTER

151

To use Cevas theorem, we will need to calculate the ratio Ia B 00 /Ic B 00 . Recalling
that CB 0 = AB 0 (since B 0 is the midpoint of AC), we can combine the four equations 7.2 and 7.3 and cancel out the equal terms to obtain:
sin2
Ia B 00
=
.
Ic B 00
sin2

(7.4)

Using completely symmetric methods, and labeling Ia Ib Ic = we get:

Ic A00
sin2
=
Ib A00
sin2

and

Ib C 00
sin2
=
Ia C 00
sin2

(7.5)

Combining the equations 7.4 and 7.5 yields:

Ia B 00 Ic A00 Ib C 00
sin2 sin2 sin2
= 1,
=
00
00
00
Ic B Ib A Ia C
sin2 sin2 sin2
which, together with Cevas theorem, proves the concurrence of the three lines at the
mittenpunkt.

7.9 A New Center

A

B
C

B
A

Figure 7.13: A New Center

Tricenters/Tcenter.T [M]

It is quite easy to find additional centers. With just a little experimentation

using geometer, the following construction appears to generate a triangle center. See figure 7.13.
Given a triangle, construct what is
sometimes called its superior triangle by drawing lines through each vertex parallel to the opposite side. This
will produce an inverted triangle similar
to the original with four times the area
and in fact, apparently composed of four

congruent copies of the original triangle.

In each of the three outer triangles, construct the incircle and connect the three
centers of those incircles to the opposite vertex of the original triangle. Those three
lines appear to meet at a point. Do they? Testing the diagram with Geometer seems
to indicate that they do. Unfortunately, other than indicate that the lines seem to be
concurrent, Geometer does not give much help toward finding a proof.

7.9.1 Geometer Analysis of the Problem

Try drawing some additional line segments and run Geometers test again. After a few
tries, one can find an interesting set of segments: connect A and B with C 00 , B and C
with A00 and A and C with B 00 . With these additions, Geometers test of the diagram
yields the following interesting relationships:

152

CHAPTER 7. TRIANGLE CENTERS

Parallel lines:
(BA) (AB)
Parallel lines:
(BC) (CB)
Parallel lines:
(AC) (CA)
Equal length segments:
[B A] [A B]
Equal length segments:
[B C] [C B]
Equal length segments:
[A C] [C A]

If these are true, it is clear that AC 00 A00 C, BC 00 B 00 C and AB 00 A00 B are all parallelograms whose diagonals bisect each other. Since every diagonal bisects the other,
all three must meet at a point.
Since all three exterior triangles are congruent, it is clear that the lengths in the list
above are equal. That they are parallel is also clear since, for example, BC 00 and CB 00
are angle bisectors of corresponding angles in congruent triangles that have parallel
sides.
This is a very rough sketch of the proof, but it is not difficult to fill in the details.

7.10 Additional Triangle Centers

There are, in fact, an infinite number of triangle centers, many of which, together with
their properties, are listed in [Kimberling, 1994]. Without proof, here is a list of a few
more of them, described in geometric terms.
The Lemoine Point. If each of the medians of a triangle is reflected across the
corresponding angle bisector, then those three reflections are coincident at the
Lemoine point.
The Spieker Center. The existence of the Spieker center is obviousit is just
the incircle of the triangle whose vertices are the midpoints of the sides of the
original triangle (called the medial triangle). Physically, it is the center of mass
of the perimeter of the triangle. In other words, if a uniform wire were bent into
the shape of the triangle, it would balance at the Spieker center.
The Feuerbach Point. This is the point of tangency of the incircle of a triangle
with its nine-point circle. The existence of this point is proved by Feuerbachs
theorem. See Section 8.17. The Feuerbach point is not precisely a triangle center,
since it fails to exist for an equilateral triangle. When the triangle is equilateral,
the nine-point circle and the incircle are the same so there are an infinite number
of points of tangency.
It is quite easy, especially with a computer geometry program, to experiment with
various symmetric combinations of lines and to find other possible triangle centers.
Often these centers turn out to be one of the many known centers and if that is the case,
it is usually easy to prove that the new center makes sense.

7.11. TRILINEAR COORDINATES

153

7.11 Trilinear Coordinates

Trilinear coordinates provide a method to locate points on the plane relative to some
given fixed triangle 4ABC (having sides a, b and c opposite the corresponding angles). The trilinear coordinates of a point are just a ratio of the (directed) distances
from the point to the sides of the triangle. If only points inside the triangle are considered there is no need for the directed in the definition, but in fact trilinear coordinates
can be used to identify any point of the plane, be it inside, outside, or on the triangle.
The directed in directed distance means that the distance is positive if the point is
on the same side of the edge as the triangle and negative otherwise 2 .
The trilinear coordinates of a point are written : : where , and are
three real numbers, at least one of which is non-zero. The three numbers are proportional to the directed distances of the point from the sides a, b and c, respectively.
The trilinear coordinates of the incenter are 1 : 1 : 1 since the incenter is equally
far from each of the sides, and it is inside the triangle. The trilinear coordinates of the
three excenters are similar: 1 : 1 : 1, 1 : 1 : 1 and 1 : 1 : 1 since those points are
also equidistant from the three edges, but they are outside the triangle in one of three
directions.
The trilinear coordinates of a point are not unique. If : : correspond to
a point, then so do : : , where is any non-zero real number. Thus the
coordinates of the incenter could equally well have been written: 2 : 2 : 2, 1 : 1 :
1, 2/3 : 2/3 : 2/3, or even : : . This is not such a bizarre concept; we
already do something similar with fractions. Although 3/5 is usually not expressed
that way, we all agree that it is equal to 6/10 or even (3)/(5). Coordinates that
are unique only up to a real multiple are said to be homogeneous coordinates, so
trilinear coordinates are sometimes referred to as homogeneous trilinear coordinates.
At least one of the coordinates must be non-zero, but up to two of them may be
zero. The point A has trilinear coordinates 1 : 0 : 0 (or : 0 : 0), and any point on the
line b has trilinear coordinates : 0 : since it is a distance 0 from the line b but can
be anywhere relative to the other two lines.
As one non-trivial example, consider the centroid of a triangle (the intersection of
the three medians). The trilinear coordinates of the centroid are 1/a : 1/b : 1/c. This
is not too hard to see: Remember that the medians divide the triangle into six smaller
triangles of equal area, so any two of them will amount to 1/3 the total triangle area.
If we drop perpendiculars to each of the sides of the triangle, these can be thought of
as altitudes of triangles 4ABG, 4BCG and 4CAG, where G is the centroid. If the
altitudes to sides a, b and c have lengths ha , hb and hc then from the equality of the
areas we obtain: aha = bhb = chc . Thus the values of ha , hb and hc (which are the
distances to the sides) are proportional to 1/a : 1/b : 1/c, which is what we wished to
show.
2 This is not exactly correct, since we will allow multiplication of all the coordinates by an arbitrary
value, positive or negative. A point inside the triangle will have all its directed distances positive, but if all
are multiplied by 1 they will all be negative and no such point exists on the plane. If a set of trilinear
coordinates is nonsensical, flip the signs on all of them to obtain a set that makes geometric sense.

CHAPTER 7. TRIANGLE CENTERS

154
Incenter
Excenters
Centroid
Circumcenter
Orthocenter
Nine-point center
Gergonne point
Nagel point
Mittenpunkt
Spieker center
Feuerbach point
Lemoine point
Fermat point
1st Napoleon point

1:1:1
1 : 1 : 1, 1 : 1 : 1 and 1 : 1 : 1
csc A : csc B : csc C
1/a : 1/b : 1/c
cos A : cos B : cos C
a(b2 + c2 a2 ) : b(c2 + a2 b2 ) : c(a2 + b2 c2 )
sec A : sec B : sec C
cos(B C) : cos(C A) : cos(A B)
sec2 (A/2) : sec2 (B/2) : sec2 (C/2)
1/a(b + c a) : 1/b(c + a b) : 1/c(a + b c)
csc2 (A/2) : csc2 (B/2) : csc2 (C/2)
(b + c a)/a : (c + a b)/b : (a + b c)/c
cot(A/2) : cot(B/2) : cot(C/2)
b+ca:c+ab:a+bc
(b + c)/a : (a + c)/b : (a + b)/c
1 cos(B C) : 1 cos(C A) : 1 cos(A B)
a:b:c
csc(A + 60 ) : csc(B + 60 ) : csc(C + 60 )
csc(A + 30 ) : csc(B + 30 ) : csc(C + 30 )

Table 7.1: Some Trilinear Coordinates

It is an interesting exercise to determine sets of trilinear coordinates for other special points. Table 7.1 lists the trilinear coordinates for all the circle centers discussed
in this chapter. Sometimes two versions of the coordinates are listed, and when that is
the case, both are equally valid (one set is a multiple of the other). It is not a difficult
exercise to see why the trilinear coordinates for the centroid can written equally well
as 1/a : 1/b : 1/c or csc A : csc B : csc C. Do you think that the 60 and the 30
in the trilinear coordinates of the Napoleon and Fermat points might have something
to do with the fact that they are constructed by erecting isosceles triangles with base
angles of 60 and 30 ? It might be worth investigating.

7.12 Triangle Center Finder

Since virtually all the triangle centers can be expressed in terms of their trilinear coordinates as displayed in Table 7.1 we can make a Geometer diagram that will help
us discover if some triangle center we find is the same as one that has already been
discovered.
The diagram Tricenters/Finder.T does just that.
It uses trilinear coordinates for an arbitrary 4ABC to draw many of the more
common triangle centers on that triangle, but on different layers.
None of the centers is drawn on layer zero, so to use this diagram, load it into
Geometer and draw your newly-discovered triangle center on layer zero. Draw the

7.12. TRIANGLE CENTER FINDER

155

final center in a different color (lets say youve made it cyan). Now edit the Geometer
code and find your point. It will be near the end of the file (probably the very last line
in the file, in fact), and it will contain text that looks something like this:
v27 = .v.ll(l12, l41, [.cyan, .in]);

(This assumes you made the final point as the intersection of two linesall that is
really important is this part: [.cyan, .in] which says that the point is painted cyan on
layer zero and is invisible on all the others. Change [.cyan, .in] to .cyan, and
exit the editor. Now your point will be visible on all layers.
To see which triangle center it is (if any), press the Next button until your point
seems to line up with one of the presented points. Then, to make sure, move the
triangle vertices around and make sure that your point stays locked on that center. If
not, continue the search.
Of course not all centers are listedjust a few of the more common ones. For
a much more exhaustive list, see [Kimberling, 1994]. Almost any of Kimberlings
centers could be added to this diagram, and he lists more than 100 of them.
Heres an exercise to test the method. Try to figure out which triangle center this
is: Construct the medians of a triangle, but reflect each one across the angle bisector of
the angle from which it emanates. Those three reflections of the median lines are also
concurrent at a point. (In fact, the point is the isogonal conjugate of the median, but it
has a special name. What is it?)
Here is some of the Geometer code for the triangle center locator. See the file on
the CD for a complete listing.
.geometry "version 0.40";
.l0;
.text("Triangle Center Locator: On this figure,
construct the center that interests you. Make
the final point non-white (cyan, for example).
Edit the geometry, and you will find toward the
end your cyan point containing [.cyan, .in].
Change this to .cyan. Then step through
the proof", .l0);
.macro .vertex trilinear(.vertex v1, .vertex v2, .vertex v3,
.flt f1, .flt f2, .flt f3)
{
ang1 = .a.vvv(v3, v1, v2, .in);
ang2 = .a.vvv(v1, v2, v3, .in);
ang3 = .a.vvv(v2, v3, v1, .in);
a1 = .f.rpn(f2, ang1, .sin, .mul, f3,
f2, ang1, .cos, .mul, .add,
.atan2);
a2 = .f.rpn(f3, ang2, .sin, .mul, f1,
f3, ang2, .cos, .mul, .add,
.atan2);
a3 = .f.rpn(f1, ang3, .sin, .mul, f2,
f1, ang3, .cos, .mul, .add,
.atan2);
A1 = .a.f(a1);

CHAPTER 7. TRIANGLE CENTERS

156
A2 = .a.f(a2);
A3 = .a.f(a3);
va1 = .v.avv(A1, v3, v1, .in);
va2 = .v.avv(A2, v1, v2, .in);
va3 = .v.avv(A3, v2, v3, .in);
l1 = .l.vv(v1, va1, .in);
l2 = .l.vv(v2, va2, .in);
l3 = .l.vv(v3, va3, .in);
.return v4 = .v.ll(l1, l2, .in);

}
v1 = .free(-0.305389, 0.511976, "A");
v2 = .free(0.691617, -0.45509, "B");
v3 = .free(-0.811377, -0.236527, "C");
a1 = .a.vvv(v3, v1, v2, .in, "\alpha");
a2 = .a.vvv(v1, v2, v3, .in, "\beta");
a3 = .a.vvv(v2, v3, v1, .in, "\gamma");
a = .f.vv(v2, v3);
b = .f.vv(v1, v3);
c = .f.vv(v1, v2);
incenter = trilinear(v1, v2, v3, 1.00, 1.00, 1.00, .l1, "I");
l1 = .l.vv(v1, v2);
l2 = .l.vv(v2, v3);
l3 = .l.vv(v3, v1);
excenter1 = trilinear(v1, v2, v3, -1.00, 1.00, 1.00, .l1, "E\sub{1}");
excenter2 = trilinear(v1, v2, v3, 1.00, -1.00, 1.00, .l1, "E\sub{2}");
excenter3 = trilinear(v1, v2, v3, 1.00, 1.00, -1.00, .l1, "E\sub{3}");
ainv = .f.rpn(1.000000, a, .div);
binv = .f.rpn(1.000000, b, .div);
cinv = .f.rpn(1.000000, c, .div);
centroid = trilinear(v1, v2, v3, ainv, binv, cinv, .l1, "M");
circa = .f.rpn(a1, .cos);
circb = .f.rpn(a2, .cos);
circc = .f.rpn(a3, .cos);
circumcenter = trilinear(v1, v2, v3, circa, circb, circc, .l1, "O");
ortha = .f.rpn(1.000000, circa, .div);
orthb = .f.rpn(1.000000, circb, .div);
orthc = .f.rpn(1.000000, circc, .div);
orthocenter = trilinear(v1, v2, v3, ortha, orthb, orthc, .l1, "H");
v4 = .v.vvmid(circumcenter, orthocenter, .l1, "N");
.text("Elementary Centers:
I, E\sub{1}, E\sub{2}, E\sub{3}: Incenter and 3 Excenters
O, M, H: Circumcenter, Centroid, Orthocenter
N: Nine-point center.", .l1);
lemoine = trilinear(v1, v2, v3, a, b, c, .l2, "L");
gerg1 = .f.rpn(1.000000, a1, 0.500000, .mul, .cos,
.dup, .mul, .div);
gerg2 = .f.rpn(1.000000, a2, 0.500000, .mul, .cos,
.dup, .mul, .div);
gerg3 = .f.rpn(1.000000, a3, 0.500000, .mul, .cos,
.dup, .mul, .div);
gergonne = trilinear(v1, v2, v3, gerg1, gerg2, gerg3, .l2, "G");
nag1 = .f.rpn(b, c, .add, a, .sub,
a, .div);
nag2 = .f.rpn(c, a, .add, b, .sub,
b, .div);
nag3 = .f.rpn(a, b, .add, c, .sub,
c, .div);

7.13. SEARCHING FOR THE TRUE CENTER OF A TRIANGLE

157

nagel = trilinear(v1, v2, v3, nag1, nag2, nag3, .l2, "N");

mit1 = .f.rpn(b, c, .add, a, .sub);
mit2 = .f.rpn(c, a, .add, b, .sub);
mit3 = .f.rpn(a, b, .add, c, .sub);
mittenpunkt = trilinear(v1, v2, v3, mit1, mit2, mit3, .l2, "M");
.text("More centers:
G, L: Gergonne Point, Lemoine Point
N, M: Nagel Point, Mittenpunkt.", .l2);

Although there is quite a bit of it (and even more in the actual Geometer file), the
structure is not too complex. First, a macro is defined that calculates the location of a
point given the three vertices of the triangle and the three trilinear coordinates. Later
we will discuss how it works.
After the macro is defined, a long series of points are drawn by passing the macro
the appropriate trilinear coordinates. For something as simple as the incenter with
trilinear coordinates 1 : 1 : 1, the code simply calls the macro with the vertices A, B
and C of the triangle together with 1, 1 and 1. The output of the macro is labeled to
indicate which center it is.
After the incenter and the three excenters is the generation of the centroid with
trilinear coordinates 1/a : 1/b : 1/c. First is a calculation (using the rpn calculator
inside Geometer) of the values of 1/a, 1/b and 1/c followed by a call on the macro.
The same procedure is followed for points with more complex descriptions, but
although the calculations are a bit messier, the idea is the same.
Now, how does the trilinear macro itself work? It is based on the idea that the
locus of points whose distance to two fixed lines have a constant ratio is a straight line
passing through the intersection of the lines. If we can find a single additional point
on the line, we can draw it. These lines of constant ratio corresponding to the trilinear
coordinates are drawn from two vertices of the triangle, and their intersection must
match the trilinear ratios with all three lines.

7.13 Searching for the True Center of a Triangle

We have learned that unless you are speaking about an equilateral triangle, there are
many different points that can be considered to be the center of a triangle, and depending upon your application, there are times when each of them is the one that makes
sense.
But here is an interesting topic to investigate that will do two things. First, it will
teach you something about geometry, and second, if you want, it can teach you about
some of Geometers more advanced features, particularly macros 3 .
3 This problem was inspired by a wonderful lecture given in about 1968 by Richard Feynman who wanted
to find the true average of two numbers. He took the arithmetic mean and the geometric mean, but since
those are not necessarily the same he repeated the process with those two means. The pair of numbers obtained in this manner converges extremely rapidly, but unfortunately to a number that can only be expressed
as a complete elliptic integral.

158

CHAPTER 7. TRIANGLE CENTERS

Here is the problem: We know that in general, the centroid, nine point center,
and incenter of a triangle are different points. But if there is a true center of the
triangle, perhaps they are close to it. So form the triangle from these three points
and find its true center. Since we do not know where the true center is, we can
perhaps approximate it by finding the centroid, nine point center, and incenter of this
new triangle. Repeat the process. Do the sets of three points tend to a limit inside
the triangle? Where is it? What if we had used different centers, like the centroid,
orthocenter, and Fermat point?
We can investigate the convergence of these sets of points by using Geometer to
repeatedly calculate the three centers, then the three centers of the triangle formed by
those three centers, and so on. Unfortunately, if the centers do get close together, it will
become harder and harder to do the construction.
The solution is to write Geometer macros and then to call those repeatedly to see
if the points converge. The file Tricenters/TrueCenter.T does just this. Here is a
listing:
.geometry "version 0.60";
v1 = .free(-0.320359, -0.227545, "A");
v2 = .free(0.185629, 0.314371, "B");
v3 = .free(0.0479042, -0.505988, "C");
.macro .vertex centroid(.vertex v1, .vertex v2, .vertex v3)
{
v4 = .v.vvmid(v1, v2, .in);
v6 = .v.vvmid(v3, v1, .in);
l1 = .l.vv(v2, v6, .in);
l2 = .l.vv(v3, v4, .in);
.return v7 = .v.ll(l1, l2);
}
.macro .vertex ninepointcenter(.vertex v1, .vertex v2, .vertex v3)
{
v12 = .v.vvmid(v1, v2, .in);
v23 = .v.vvmid(v2, v3, .in);
v31 = .v.vvmid(v3, v1, .in);
c1 = .c.vvv(v12, v23, v31, .in);
.return v4 = .v.ccenter(c1);
}
.macro .vertex incenter(.vertex v1, .vertex v2, .vertex v3)
{
l1 = .l.vv(v1, v2, .in);
l2 = .l.vv(v2, v3, .in);
l3 = .l.vv(v3, v1, .in);
c1 = .c.lll(l1, l2, l3, 1, .in);
.return v4 = .v.ccenter(c1);
}
vg0 = centroid(v1, v2, v3, .magenta);
vo0 = ninepointcenter(v1, v2, v3, .magenta);
vi0 = incenter(v1, v2, v3, .magenta);
vg1 = centroid(vg0, vo0, vi0, .red);
vo1 = ninepointcenter(vg0, vo0, vi0, .red);
vi1 = incenter(vg0, vo0, vi0, .red);
vg2 = centroid(vg1, vo1, vi1, .yellow);
vo2 = ninepointcenter(vg1, vo1, vi1, .yellow);
vi2 = incenter(vg1, vo1, vi1, .yellow);

7.13. SEARCHING FOR THE TRUE CENTER OF A TRIANGLE

159

The diagram begins with the three points of the original triangle drawn in white.
Next are three macros to construct the three centers that interest us. The vertices vg0,
vo0 and vi0 are the centroid, nine point center and incenter of that triangle. Replace
the 0s with 1s and we obtain the second level of centers, et cetera. Each set is drawn in
a different color to see if the points converge or not. Experiment with the diagram and
see what you can learn. See what happens if you go another step or two deeper.
Sometimes bad things happen. Which ones are due to round-off error in the computer, and which are due to features of the original triangle?
If you wish to experiment with different triangle centers, the easiest way to build a
macro is this: In the original triangle using the graphical user interface, construct the
center in the normal way. For example, if you are just beginning and want to build the
macro to draw the incenter, begin with a diagram having three free points and construct
the incenter. Your Geometer diagram will look something like this:
.geometry "version 0.60";
v1 = .free(-0.320359, -0.227545, "A");
v2 = .free(0.185629, 0.314371, "B");
v3 = .free(0.0479042, -0.505988, "C");
v4 = .v.vvmid(v1, v2, .in);
v6 = .v.vvmid(v3, v1, .in);
l1 = .l.vv(v2, v6, .in);
l2 = .l.vv(v3, v4, .in);
v7 = .v.ll(l1, l2);

In the text editor it is an easy exercise to convert this code to the macro form in
the original listing. The portion of the code above that will go into the macro includes
the line beginning with v4 = .v.vvmid and continues to the end of the listing. To test
it before going on, add a single line of code (that you may wish to erase later) so that
your entire program looks like this:
.geometry "version 0.60";
v1 = .free(-0.320359, -0.227545, "A");
v2 = .free(0.185629, 0.314371, "B");
v3 = .free(0.0479042, -0.505988, "C");
.macro .vertex centroid(.vertex v1, .vertex v2, .vertex v3)
{
v4 = .v.vvmid(v1, v2, .in);
v6 = .v.vvmid(v3, v1, .in);
l1 = .l.vv(v2, v6, .in);
l2 = .l.vv(v3, v4, .in);
.return v7 = .v.ll(l1, l2);
}
testvertex = centroid(v1, v2, v3, .magenta);

If it does the right thing, save your file and start working on the definition of the
macro for the second triangle center.

160

CHAPTER 7. TRIANGLE CENTERS

Chapter

Inversion in a Circle
Four circles to the kissing come,
The smaller are the better.
The bend is just the inverse of
The distance from the centre.
Though their intrigue left Euclid dumb
Theres now no need for rule of thumb.
Since zero bends a dead straight line
And concave bends have minus sign,
The sum of squares of all four bends
Is half the square of their sum1 .
1 This poem is the Descartes circle theorem, not proved in this book. If r = 1/a, r = 1/b, and
1
2
r3 = 1/c are the radii of three mutually tangent circles, then the radius r 4 = 1/d of the circles tangent
to all three satisfies the equation 2(a2 + b2 + c2 + d2 ) = (a + b + c + d)2 There are two solutions
to the equation corresponding to the two circles tangent to the three given circles. The negative solution
corresponds to the surrounding circle and the positive one to the surrounded circle. See figure 8.1

161

CHAPTER 8. INVERSION IN A CIRCLE

162
Frederick Soddy

8.1 Overview of Inversion

Inversion in a circle is an powerful tool that can be applied
to many problems in Euclidean geometry. Geometer has
built-in commands supporting inversion in a circle. Any
time you are faced with the construction of circles that are
tangent to a combination of lines and circles, it is wise
to consider inversion as a possible technique for finding a
solution.
Figure 8.1: Soddy Circles
In normal usage, the word inversion means turning
Inversion/Soddy.D [D]
inside-out, and this is close to the technical geometrical
meaning. The key idea is this: if some figure is inverted
(turned inside-out) and then the result is inverted again, the final figure will be identical
to the original.
The technique of inversion is very much like changing coordinate systems to simplify a problem. In computer graphics, for example, there is a very simple equation to
calculate the position of an object after it is rotated about the origin, but the equations
to calculate rotation about an arbitrary point in space are fairly ugly. Thus, to perform
rotation about an arbitrary point, the usual technique is to translate all the coordinates
so that the point of rotation is at the origin. The desired rotation is then applied to
objects expressed in these new coordinates, and finally, the result is translated back by
undoing the translation (in other words, by translating in the opposite direction).
Inversion turns the points of the plane inside-out relative to some circleall the
points that were inside the circle wind up on the outside and vice-versa. The points on
the circle of inversion itself are not moved. Inversion is a sort of reflection through a
circle.
Although inversion is defined simply in terms of its effects on points, what is far
more interesting is what inversion does to lines and circles. To invert a line, for example, invert every point on the line through the circle. The resulting set of points is the
inverse of the line. The same idea applies to circles.
We will show that the inverted image of a line or circle is another line or circle.
By carefully choosing the circle of inversion, you can control when a line is inverted to
become a circle, and when it remains a line. Similarly, you can invert circles to become
lines, to become other circles, or even make them invert into themselves.
Inversion also supplies the option of sending troublesome points to the point at
infinity where they are so far away that they no longer cause problems.

8.1.1 Inversion of Points

Inversion is always relative to a circle in the plane. In this overview, we will invert
through a circle K with center O. Inversion in the circle K (or relative to K) is an

8.1. OVERVIEW OF INVERSION

163

operation on the points of the plane that moves almost every point to a new location. If
the point was outside K, its inverse will be inside, and vice-versa. If the point is on the
circle K inversion leaves it unchanged (moves it to itself).
Inversion always moves a point directly toward or away from the center of the circle K.
In other words, if you draw a ray connecting the
center O of K through the point P , the inverse of

Figure 8.2: Inversion in a Circle

Inversion/Inversion.T [M]

P will lie on the same ray OP . If K has radius

r = OR, and if point P is inverted through K to
point P 0 , then the formula relating the lengths is
this (see figure 8.2:
OR
OP
=
,
OR
OP 0

or OP OP 0 = r2 .

(8.1)

If you invert a point P in circle K to obtain a point P 0 , then if you invert P 0 in K,

the result will be the original point P . In other words, repeated inversions in the same
circle take a point back and forth between two positions. Another way to think about it
is that a second identical inversion undoes the first. This is clear from equation 8.1.
Mathematically, we say that the operation of inversion in a circle is self-inverse.
Inversion is much like the operation of inversion in the real numbers that takes a
number x and converts it to the number 1/x. Inverting twice in the real numbers leaves
the number unchanged2 (2 1/2 1/(1/2) = 2, or x 1/x 1/(1/x) = x).

8.1.2  Inversion of the Circle Center

Inversion has the property that points near the edge of K are inverted to points also near
the edge, but on the other side. Points close to the edge of K, but inside K are inverted
to points that are also close to the edge of K, but on the outside, and vice-versa.
Conversely, points very near the center of K (in other words, almost as far from the
edge as possible on the inside) are mapped to points very far away from the circle. In
fact, if you move a point gradually toward the center of the circle K of inversion, its
inverse image moves more and more rapidly away from the center, or toward infinity.
In the Euclidean plane, the point at the center of K cannot be invertedit would
have to go to a point infinitely far from the center. If you imagine adding a point
at infinity to the Euclidean plane, you can make a mathematically sensible definition
for inversion that includes this point at infinity. The inverse of the center of K is
the point at infinity, and the inverse of the point at infinity is the center of the circle of
inversion.
2 Inversion in the real numbers is actually very similar to geometric inversion in a circle. Numbers inside
the range (1, 1) are inverted to numbers outside that range, 1 and 1 are fixed, and geometric inversion of
the center of the circle has exactly the same problem as inversion in the line has with the number zero.

164

CHAPTER 8. INVERSION IN A CIRCLE

8.1.3 Inversion of Shapes

To invert a shape such as a line or a circle, all that is required is to invert every one of
its points. We will show that the inverses of lines and circles are also (almost) lines and
circles, but at first it is a bit confusing because inverted lines can become either lines
or circles, and similarly for inverted circles.
It is often fruitful to think about this by lumping lines and circles together, and just
think of a line as an infinitely large circle. In the rest of this paragraph, quotation
marks around the word circle will indicate that the word means either a circle or a
straight line. This makes some sense, since small parts of very big circles begin to
look more and more like straight lines. Since this is the case, why not just think of
straight lines as special circles that have an infinitely large radius? If you think of it
that way, then the theorems about the inversion of lines and circles all look the same.
If you invert a circle in K, you get another circle3 .

8.2 Formal Definition of Inversion

The definition of what it means to invert a point in a circle is simple, but to get an intuitive feeling for inversion, it is a good idea to experiment with the Geometer diagram
for figure 8.3 that allows you to move the point P and see how its inverse moves in
response. Here are some ideas: Try moving P toward the center and away. Place P on
the edge of the circle. Try to move P along various straight lines or circles and watch
the path of P 0 .
Given a circle with center O and a point P as
R
in figure 8.3, if P is outside the circle, construct
a tangent to the circle at R through P , connect
P and O, and then drop a perpendicular from R
to P O, intersecting P O at a point P 0 . This P 0 is
P
P
O
the inverse of the point P relative to that circle.
To find the inverse if P is inside the circle, do
just the opposite: connect O and P with a line,
Figure 8.3: Inversion in a Circle
construct a perpendicular to OP at P that hits the
Inversion/Inversion.T [M]
circle at R, then construct a tangent to the circle
through R which will hit the line OP at a point
P 0 . Clearly if P 0 is the inverse of P relative to some circle, then P is the inverse of P 0
relative to the same circle.
There are two other cases to consider where the construction above fails. What if
P is on the circle, and what if P is at the center of the circle?
If you experimented with a computer geometry diagram, the answer to the first
question should be obviousthe closer P is to the edge of the circle, the closer P 0 is
3 In fact, there is no reason that K itself needs to be a circleyou can invert in any circle, even if it is
a straight line. Inversion in a line (thought of as an infinitely large circle) turns out to be exactly the same as
a mirror reflection of the points across the line.

8.3. SIMPLE PROPERTIES OF INVERSION

165

to P . It only makes sense to define the inverse of a point on the circle to be the point
itself, so that is how it is defined.
If P is at O, the center of the circle, it is another story. The closer you move P
to O, the further away P 0 moves. If you look at what happens in our geometrical
construction described above, you will see that the closer P is to O, the closer RP is
to being parallel to OP . When P is at O, the construction will yield a pair of parallel
lines. For this reason, we will simply say that the inverse of the center of the circle is
undefined.
Undefined is used in the same sense as in arithmetic when we say that 1 divided
by 0 is undefinedthere is simply no sensible way to define 1/0 so that it makes
mathematical sense. In the same way that even though it is undefined, you can think
of 1/0 as being (after all, 1/ is very large if  is very small), you can think of the
inverse of the center of the circle as being a point at infinity. You cannot use this
idea in a formal proof, but nobody can prevent you from thinking about it like that, and
thinking about it like that can sometimes be very useful.
We only have to make the point undefined in pure Euclidean geometry where we
are only allowed to talk about points in the Euclidean plane. If we are willing to make
a new kind of inversive geometry where we add a point at infinity to the Euclidean
plane, we can formally define inversion of the center of the circle. For now, let us not
change geometries in mid-chapter.

8.3 Simple Properties of Inversion

In this section, assume that all inversions will be relative to a circle called K with
center O and radius r.

8.3.1 Inversion Expressed as a Ratio

We can express inversion as a ratio. Since the radius of the circle is r, what is the
mathematical relationship between the lengths OP 0 and OP in figure 8.3? There are
three similar triangles: 4ORP 4OP 0 R 4P P 0 R. Using the similarity of the
first and last triangles, we obtain:
OP
OR
=
OR
OP 0
or, as it is more commonly written:
OP OP 0 = OR OR = r 2 .

(8.2)

In many books equation 8.2 is used as the definition of inversion. The construction we
used to define it is derived as a result. Both definitions are equivalent.
This equation shows that the concept of the inverse of the center as a point at
infinity is exactly equivalent to the concept that 1/0 = as OP gets close to zero,
OP 0 gets close to r 2 /0 = .

CHAPTER 8. INVERSION IN A CIRCLE

166

8.3.2  Inversion in the Complex Plane

There is a nice analytic definition of inversion if you know something about complex
numbers (see Section A.13).
If o is an imaginary number somewhere on the complex plane, and r is a real radius,
then the transformation that inverts all the complex numbers z to their complementary
points z 0 through a circle of radius r centered at o in the complex plane is given by:
z0 =

r2
+ o.
z o

This clearly has the right propertiesif z is very close to o (the center of inversion),
then z will be very close to o, so the denominator is very small and the result will be
very large. Similarly, if we move the o to the other side and multiply both sides by
(
z o) to get rid of the fraction, we obtain:
(z 0 o)(
z o) = r 2

which looks a lot like the geometric definition, since the absolute value of z 0 o represents the distance between z 0 and o.
If inversion in the complex plane is relative to the origin (in other words, if o = 0),
and if the radius of the circle of inversion is 1 (in other words, if inversion is relative to
the unit circle), then the inverse of a (complex) point z is 1/
zvery similar to inversion
in the real line, since if z happens to be real, z = z.

8.3.3 The Inverse of a Line

Inverting single points relative to a circle is not too interesting, but what if all the points
on a line are inverted? What will be the resulting shape? Using Geometer, it is very
easy to run experiments like thistake a point that is constrained to be on a line, and
invert it through a circle. Then move the point along the line and see which points are
swept out by its inverse.
A bit of experimentation should convince you that in almost every case, the inverse
of a line seems to be a circle. Can you find the case where it is not? Let us prove the
following theorem:
Theorem 8.1 (Inverse of a Line) If K is a circle and L is a line that does not pass
through the center of K, then the locus of the inverses of the points on L is a circle
that passes through the center of K. If L lies completely outside of K, its inverse will
be completely inside K. If L is tangent to K, its inverse will be a circle inside K, but
tangent to K at the point where L is tangent. If L intersects K at two points, its inverse
will be a circle intersecting K in those same two points. The inverse of L is not quite
the entire circleit is the circle missing the point that is the center of K. If L passes
through the center of K, the points on the inverse of L lie on L itself (although the
points of L that were inside K map to points outside, and vice-versa). Of course if L
passes through the center of K, that center point can not be inverted, and the inverse
of L will be L itself with the exception of the point at the center of the circle.

8.3. SIMPLE PROPERTIES OF INVERSION

167

It is easy to construct the circle that is the

inverse of L. For the case where L does not
intersect K, see figure 8.4.
Let O be the center of circle K, and from
O, drop a perpendicular line to L, meeting
L at U . Let U 0 be the inverse of U relative
to the circle K. We will show that the circle
with diameter OU 0 is the inverse of line L.

R
P
U

P
U

To prove this, let P be any point on L, Figure 8.4: Invert a Line in a Circle
and construct the inverse P 0 of P . If r is the (Case 1)
radius of K, we have:
Inversion/Inversion1.T [M]
OP OP 0 = r2 = OU OU 0
so

OU
OP
=
.
OU 0
OP 0

Since they share the angle at O, 4OP 0 U 0 4OU P , and since OU P = 90 ,

we have OP 0 U 0 = 90 . For every P , OP 0 U 0 = 90 , so all possible points P 0 must
lie on the circle with diameter OU 0 .
The second case, where L is tangent to K, is simple. Exactly the same diagram
can be used, except that U = U 0 , so the argument works for every point except where
U = U 0 , but in that case, the inverse of U , since it lies on the circle, is U = U 0 itself,
and U 0 is one end of the circles diameter, so it is on the circle as well. The inverse of a
line that is tangent to the circle of inversion is a circle passing through the center of the
circle of inversion and tangent to the circle of inversion at the same point that the line
is tangent. In other words, it is a circle exactly half the size of the circle of inversion,
and it fits snugly between the center and edge of that circle.
In the third case, L passes through K in
two points (see figure 8.5). The argument is
virtually identical, except when we drop the
perpendicular from O to L, the intersection,
U , will be inside the circle. Thus the inverse, U 0 will be outside, and the image circle, which will still be the circle with diameter OU 0 , will pass through the circle in two
points which will coincide with the intersections of L and K.

P
U
O
Q

In this situation, there are a few cases to

consider. The points on the line that we are
inverting can be either inside, outside, or on Figure 8.5: Invert a Line in a Circle
circle K. If they are on circle K, they are (Case 3)
their own inverses, so points where L passes
Inversion/Inversion2.T [M]
through K are fixed under inversion. If the
point is inside (look at P , for example, in figure 8.5), it is easy to show by the inversion

CHAPTER 8. INVERSION IN A CIRCLE

168

ratios that 4OP 0 U 0 4OU P , so OP 0 U 0 = 90 , so P 0 lies on the circle with diameter OU 0 . If the point on L is outside circle K (see point Q, for example), then since
OU/OQ = OQ0 /OU 0 and since they have the included angle QOU in common, we
have 4OU Q 4OQ0 U 0 . OU Q = 90 , so OQ0 U 0 = 90 , so Q0 is on the circle
with diameter OU 0 which is what we wanted to prove.
The final case where L passes through O is trivially truethe inverses of points of
L are guaranteed to lie on L by definition.

8.3.4 Inversion of a Circle in a Circle

If P inverts to P 0 in circle K then P 0 inverts to P in that same circle. From this and the
previous section, we know that the inversion of a circle that passes through the center of
K of inversion is a straight line, with the same three casesif the circle is completely
within K, the line will be completely outside it. If the circle is tangent to K, the line
will be tangent to K at the same point, and finally, if the circle passes through K in
two points, the line will also pass through K in those same two points.
But what if the circle to be inverted does not pass through the center of K?
If you experiment with a Geometer diagram you can convince yourself that the
inverse of a circle not passing through the center of K is another circle. There are
again a few cases to consider but let us look at figure 8.6. In that figure, the circle
centered at X is being inverted in K, the circle with center O. We will show that if we
connect the centers O and X and invert the endpoints D and E of the diameter along
the line OX to D0 and E 0 , then the inversion of the circle centered at X is the circle
whose diameter is D 0 E 0 .
Clearly D and E go to the right
places, but how about an arbitrary point
P on the circle centered at X? If we
poke around a bit we can find some similar triangles.
From the definition of an inversion
we have:

P
D

P
D

X
E E

OP 0 OP = OD0 OD,

Figure 8.6: Invert a Circle in a Circle

or OP 0 /OD0 = OD/OP . Therefore
Inversion/Inversion3.T [M]
4OP D 4OD0 P 0 by SAS since in
addition to the ratio above, the two triangles have DOP in common.
Using exactly the same argument, but using E and E 0 instead of D and D 0 we
have 4OP 0 E 0 4OEP . By the similarity of these two triangles, we know that
P DE = D0 P 0 O and P EO = OP 0 E 0 . But OP 0 E 0 is the supplement of
E 0 P 0 P and P EO is the supplement of P ED, so P ED = E 0 P 0 P .
Since they form a triangle,
P ED + P DE + EP D = 180 .

(8.3)

8.3. SIMPLE PROPERTIES OF INVERSION

169

And since they form a straight line,

E 0 P 0 P + D0 P 0 O + D0 P 0 E 0 = 180.

(8.4)

Since P ED in equation 8.3 is equal to E 0 P 0 P in equation 8.4 and similarly for

P DE and D0 P 0 O, we conclude that EP D = 90 = D0 P 0 E 0 , so P 0 must lie
on the circle with diameter D 0 E 0 which is what we wanted to show.
Very similar arguments can be used
if the circles intersect, or even if one circle completely surrounds the other.
Do not make the mistake of thinking
that because a circle is mapped by inversion into another circle that the center is
mapped to the center, or that all the diameters are mapped to other diameters.
All you know is that every point on the
circle is mapped to other points on the
circle, and not necessarily with uniform
Figure 8.7: Non-Uniformity of Inversion
spacing. Figure 8.7 shows this clearly
Inversion/Inversion4.D [D]
the circle on the left with uniformlyspaced points that appear as small circles is inverted in the center circle to make the circle with non-uniformly-spaced points
indicated as diamonds. The center of the original circle (also a small circle) is inverted
to the diamond far to the left of the circle of diamonds. Inversion turns things insideout, so the inverse of the center may even be outside the inverse of the original circle.
The diameters along the line passing through the centers of the circle and its inverse
are mapped into each other, but except in very special cases (for example if the circle is concentric with the circle of inversion), the other diameters will almost never be
mapped to diameters.

8.3.5 Inversion with Straightedge and Compass

All the results above are theoretical, and can be very useful in constructing a proof, but
if you actually have to draw the inverse of a line or a circle, how do you do it?
If the result is a line, simply select two points on the line or circle that is to be
inverted and invert them as described in Section 8.2. Then connect those two points
with a line.
If the result is a circle, simply invert three different points, and you will be left with
three points that must lie on the required circle. The intersection of the perpendicular
bisectors of the segments connecting the points is the center of the required circle.
Of course if you are using Geometer, just use the built-in inversion commands:
VC=>V Inv, LC=>C Inv and CC=>C Inv. Since both lines and circles usually go to
circles, Geometer assumes that the result is always a circle; if the result is a line, what
is drawn is effectively an infinitely large circle.

170

CHAPTER 8. INVERSION IN A CIRCLE

8.4 Preservation of Angles Under Inversion

When we say that inversion in a circle preserves angles, exactly what is meant? It is
clear what is meant by the angle between two lines, but what is the definition of the
angle between a line and a circle, or the angle between two circles?
The sensible way to define the angle between curves (where a straight line is a
special type of curve) is to say that the angle between them is the same as the angle
between their tangent lines at the point of intersection. Thus if a pair of circles intersect
at a point P and you want to measure the angle between them, construct the tangents
to each of the circles at P and measure the angle between those tangent lines. If you
have a circle and a line, just find the tangent to the circle at the point of intersection and
measure its angle relative to the line.

8.4.1 Inversion Preserves Angles Between Lines

Suppose AB and CD are two lines in the plane that intersect at the point P . If they
are inverted in a circle with center O, we will show that the resulting figures make the
same angle with each other.
The easiest case is if O, the center of
inversion, happens to be the same as P .
Since both lines pass through the center
of inversion, they are transformed into
themselves so their images trivially have
the same angle between them.

A
Y
X

C
P
B

O
If the center of inversion is not on
one of the lines (but it may be on the
D
other), then the line that does not contain the center will be inverted to a circle passing through O. Suppose O is not
on AB. Consider the line OX passing
Figure 8.8: Invert Two Lines in a Circle
through O that is parallel to AB (see figInversion/Angles1.D [D]
ure 8.8). Under inversion, OX is transformed into itself (since it passes through the center of inversion), and AB is inverted
into a circle K passing through O. Then OX must be tangent to K, since it touches
K at O, and if it intersected K in more than one place, OX and AB would intersect,
which is impossible since they are parallel.

If O is on CD, then CD is inverted to itself, and its angle with OX is clearly

unchanged (OX is also inverted into itself). But CD makes the same angle with OX
as with AB since they are parallel, so the inverses of AB and CD make the same
angle.
Finally, if O is on neither line, then both AB and CD are inverted to circles that
meet at O, and the lines tangent to those circles at O (OX and OY in the figure) are
parallel to the original lines, so they make the same angles as did the original lines.

8.4. PRESERVATION OF ANGLES UNDER INVERSION

171

8.4.2 Inversion Preserves Angles Between Circles

If two intersecting circles are inverted, the angle between them is also preserved under inversion. The proof basically uses the result above, but there are a few cases to
considerwhether the inversion is through the point of intersection, whether it is on
one of the circles, et cetera. We will not present the exhaustive proof here, but the
following paragraph gives the general idea.
Where the original circles intersect at a point P , construct lines tangent to them
at P , and the angle between those tangent lines is the same as the angle between the
circles. Under inversion, we know the inverses of the lines go to circles that intersect
at the same angle, and these circles have to be tangent to the images of the circles we
were inverting (since they meet the line in exactly one point). Tangent circles have the
same tangent line at their point of tangency, so the inverses of the circles meet at the
same angle as the inverses of the tangent lines, which meet at the same angle as the
original circles.

8.4.3 Orthogonal Circles

Figure 8.9: A Family of Orthogonal Circles

Inversion/Orthocircs.T [S]

An interesting situation to consider is when circles (or lines, or lines and circles)
intersect at 90 . In any of these cases, the two figures are called orthogonal. In figure 8.9 the circle in the center is surrounded by a set of circles that are orthogonal to
it.
Since angles are preserved under inversion, what happens if a circle is inverted
through another circle to which it is orthogonal? The answer is that the circle is inverted
into itself. To see why, remember that the circle of inversion is unchanged by the
inversion. The circle orthogonal to it is mapped to a circle that meets it at the same
points, and remains orthogonal. Therefore it must be mapped to the same circle. (Note
that this does not mean that the points are not moved by the inversionthe ones outside
are mapped to points inside and vice-versa, but every point on that circle is inverted to
another point on the same circle.)

172

CHAPTER 8. INVERSION IN A CIRCLE

8.5 Summary of Inversion in a Circle

Here are the key facts about inversion in a circle K centered at O having radius r.
Inversion is indicated with primesthe inverse of point P is P 0 ; the inverse of circle
is 0 , et cetera. We will be a bit sloppy in dealing with the inverse of O.
1. Points inside K are inverted to points outside K and vice-versa. Points lying on
K invert to themselves. As an important special case, note that K 0 = K.
2. OP OP 0 = r2 . The points P and P 0 lie on the same ray originating at O.
3. Circles not passing through the center of K are inverted to other circles. Circles
passing through the center are inverted to lines.
4. Lines passing through O are inverted into themselves. Lines not passing through
O are inverted to circles passing through O.
5. Lines and circles that intersect in zero, one, or two places are inverted to other
lines and circles that intersect in the same number of places. A little care must be
taken to interpret this statement correctly if the intersection or tangency is at O.
For example, if two circles are tangent at O, then their inverses will be parallel
lines (that meet at infinity). If a line and a circle are tangent at O, then the
inverse of the circle will be parallel to the line (which is inverted into itself).
6. Lines or circles intersecting or tangent to K are inverted to lines or circles intersecting or tangent to K in exactly the same places.
7. If two lines, two circles or a line and a circle meet at an angle , then the inverses
of those two figures will also meet at the same angle .
8. If a circle K1 is orthogonal to K, then K10 = K1 .
In most of the examples that follow, we will choose the circle about which we
would like to do the inversion, and we can choose it to get the effect you want. For
example, suppose we are working with a circle, and we would much rather be working
with a straight line. Then choose the circle of inversion to be one whose center lies on
the circle you want to convert into a line. After inversion, the circle we wanted to avoid
will be converted to a straight line.
If you want a straight line to remain a straight line after the inversion, then you had
better invert through a circle whose center lies somewhere on that line. If you want two
straight lines to be unchanged after the inversion, then the circle of inversion must have
its center at the intersection of those two lines.
If you have three lines that do not meet in a point, and you want them all to be
straight lines after an inversion, then, unfortunately, you are out of luck, and you will
have to find a different way to solve the problem.
Remember that Geometers inversion commands are particularly powerful. If you
have a problem that you think can be approached with inversion, draw the figure, add

8.6. CIRCLE THROUGH A POINT TANGENT TO TWO CIRCLES

173

two arbitrary points, make a circle with one as the center and the other on the edge
of the circle, and invert all the objects of interest in that arbitrary circle. Then you
can move the test circle around to see how inversion relative to circles having various
centers and various radii modifies the problem. Try moving the center over all the
interesting points in the Geometer diagram to see what happens.

8.6 Circle through a Point Tangent to Two Circles

Given two circles and a point, find the circle (or circles) that are tangent to both the
given circles and pass through the point4.

P
C1
R1

C2

C1
R2
C2

Figure 8.10: Circle through a Point and Tangent to Two Circles

Inversion/Pointcirccirc.T [P]

See figure 8.10. The problem of finding a circle tangent to two other circles seems
difficult, to say nothing of the added problem of making it go through a particular point.
It is, however, not hard to construct lines tangent to one or two circles. The problem
would be much easier if we were searching for a line instead of a circle.
The final solution will be a circle passing through P and tangent to the two circles
centered at C1 and C2 . If we could do an inversion to turn that final solution into a line,
the problem might be easier.
If a circle passes through the center of its circle of inversion, it is converted to a
line, so why not invert the entire figure through a circle centered at P since the solution
has to go through P ?
But which circle? There are an infinite number of choices with any possible radius.
The answer (in this case) is that it does not matter! Pick any circle centered at P and
invert everything through it. After inversion the circles centered at C 1 and C2 become
the circles centered at C10 and C20 . The solution circle, wherever it was, is inverted to a
4 This is a simple example of what are generally known as Apollonius Problemsto construct circles
satisfying various tangency and concurrency requirements. The most difficult of Apollonius Problems is
this: Given three circles in a plane, construct the circle (or circles) simultaneously tangent to all three of
them.

CHAPTER 8. INVERSION IN A CIRCLE

174

line, but it will still be a line tangent to the inverted copies (C10 and C20 ) of the original
circles5 .
We need to find a line tangent to the circles centered at C10 and C20 , but depending
on the relationships and radii of the two original circles, there may be as many as four
solutions or as few as zero. The solution lines will be the common external and internal
tangents to those circles. If the circles intersect, there are no common internal tangents.
If one of the inverted circles is inside the other, there will be no external tangents either.
Invert those lines (however many there are) back through the original circle of inversion to obtain all the possible solutions to the original problem. In the figure there
are four common tangents to the inverted circles, so there will be four solutions to the
problem, and all are displayed in the figure.
This is an example where a computer geometry diagram is vastly more powerful
than a pencil and paper. There are dozens of configurations to try, and it would be
very valuable for you to try them yourself. Adjust the positions and radii of the circles
centered at C1 and C2 , and move the point P around to get a feeling for what the
solutions look like, and when there are zero, one, two, three, or four solutions (all five
situations can occur).

8.7 Circle Tangent to Three Circles

C1

C2

X
C3

Figure 8.11: Circle Tangent to Three Circles

Inversion/Threetancirc.T [P]
5 If you are doing this with a straightedge and compass, it will be slightly easier to choose the circle of
inversion so that in addition to having its center at the point, it is also tangent to one of the circles. Do you
see why?

8.8. PTOLEMYS THEOREM REVISITED

175

This is a very special case of Apollonius Problem6 . Suppose that you are given
three mutually tangent circles centered at C1 , C2 , and C3 as in figure 8.11. Construct
the circles that are simultaneously tangent to all three of the mutually tangent circles.
In general there will be two solutionsone outside the three circles and one in the
space enclosed by them. Since it is hard to find tangents to circles, it would be nice to
turn them into lines. We cannot turn all three into lines, but certainly two of them can
be by inverting in a circle that is centered at the point of tangency of two of them.
Let X be the point of tangency between the circles centered at C2 and C3 . Draw
an arbitrary circle with X as its center (in the figure it is the circle passing through the
point K), and invert all three of the original circles in it. The two circles tangent at its
center will be mapped to parallel lines, and those lines will continue to be tangent to
the inverted image of the circle centered at C1 which in the figure is the circle centered
at S. The two tangent lines are the roughly horizontal lines touching the circle centered
at S above and below. It is easy to find circles centered at T and U that are tangent to
the one centered at S and which are also tangent to the two lines. There is one on each
side, they have the same radius as the circle centered at S, and their midpoints lie on
the line through S and parallel to the lines that are inverses of the other two circles.
Invert the two circles back through the one centered at X and this will yield the two
solutionsa tiny one inside and a larger one outside.
There is a purely classical solution to this problem, where classical means that
it does not use inversion. The complete classical construction can be found in the
Geometer file Inversion/ClassicSoddy.T.

8.8 Ptolemys Theorem Revisited

Zvezdelina Stankova sparked this authors interest in inversion by showing
him this most wonderful proof in the
world of Ptolemys theorem using inversion.
Ptolemys theorem is the same
given any convex cyclic quadrilateral
ABCD, show that
AB CD + BC AD = AC BD.

B
K

C
C
D
D

A
Figure 8.12: Ptolemys Theorem Revisited
Inversion/Ptolemy1.T [M]

The theorem will be proven by inverting the figure about a circle K with center at A, as shown in figure 8.12.
First we need a simple lemma that is quite useful in its own right.
6 The difficult case of Apollonius problem solves this for three arbitrary circles that are not necessarily
mutually tangent. Also, see the footnote to the poem at the beginning of this chapter.

CHAPTER 8. INVERSION IN A CIRCLE

176

Lemma 8.1 (Length of an inverted segment) If the endpoints of segment BC are inverted in a circle with center A, then the length of segment B 0 C 0 is related to the length
of BC as follows:
BC AB 0
BC AC 0
B0C 0 =
=
.
AC
AB
To prove this we can use the same figure 8.12. AB 0 AB = AC 0 AC because they
are inverse points, so AB 0 /AC 0 = AC/AB. Since BAC is equal to itself, by SAS
we have 4ABC 4AC 0 B 0 so
B0C 0
AB 0
AC 0
=
=
.
BC
AC
AB
Multiply through by BC to prove the lemma.
If the entire quadrilateral ABCD is inverted through a circle centered at A with an
arbitrary radius r, we have:
AB AB 0 = AC AC 0 = AD AD0 = r2 .
Since the circle in which ABCD is inscribed passes through the center A of the
circle of inversion, its inverse is a line, so B 0 , C 0 , and D0 lie on a line, and from the
figure, it is clear that B 0 D0 = B 0 C 0 + C 0 D0 .
Replacing B 0 D0 , B 0 C 0 , and C 0 D0 by their equivalents according to Lemma 8.1, we
have:
BC r2
CD r2
BD r2
=
+
.
AD AB
AC AB
AC AD

Cancel the r2 from all terms and multiply through by AB AC AD and obtain:
AC BD = BC AD + CD AB,
which completes the proof of Ptolemys theorem.

8.9 Fermats Problem

Ptolemys theorem has an interesting extension that can be used to solve Fermats Problem:
Fermats Problem: Let A, B, and C be any three points in the plane. Find the
point X that minimizes the sum of the lengths of the segments connecting A, B, and
C to X. In other words, make AX + BX + CX as small as possible.
Ptolemys theorem tells us that if ABCD is a cyclic quadrilateral, then AB CD +
BC DA = AC BD, but what if the four points A, B, C, and D do not happen to lie
on a circle? is there some inequality that can replace the = sign?
The answer is yes. To see why, look at a slight enhancement of Ptolemys theorem:

8.9. FERMATS PROBLEM

177

Theorem 8.2 (Ptolemys Theorem Extended) If A, B, C, and D are any four points
in the plane, then
AB CD + BC DA AC BD,
and equality holds only if ABCD is a convex cyclic quadrilateral.
Suppose that ABCD is not a cyclic
B
quadrilateral as in figure 8.13. Then we
can go through exactly the same steps
C
as we did in the proof of Ptolemys theB
orem in the previous section, except it is
C
no longer the case that B 0 C 0 + C 0 D0 =
B 0 D0 that only holds if B 0 , C 0 and D0
D
lie on a line, and that happens only for
D
A
cyclic quadrilaterals. If the three points
do not lie on a straight line, the triangle inequality gives us: B 0 C 0 + C 0 D0 Figure 8.13: Ptolemys Theorem Extended
B 0 D0 . Since other than this change of
Inversion/Ptolemyext.T [M]
= to , the proofs are identical, we
have proved the extension to Ptolemys theorem.
Now we can look at Fermats Problem, which is solved by other methods in Section 7.4.
In figure 8.14, let A, B, and C be the points in question, and on side AB construct
an equilateral triangle 4ABD pointing outward7 . We claim that the intersection X of
the line DC and the circle that circumscribes 4ABD is the required point.
Let X be an arbitrary point. Then the extension to Ptolemys theorem tells us that
AX DB + BX AD DX AB,
but since 4ABC is equilateral, we know that AD = DB = AB, so we know that
AX + BX DX.

(8.5)

Add CX to both sides of equation 8.5 (since we are trying to minimize AX +

BX + CX) to obtain:
AX + BX + CX DX + CX.

(8.6)

If X happens not to be on the line DC, then DX + CX can always be made

smaller by moving X on to the line DC by the triangle inequality. Thus to minimize
DC + CX, X must lie on the line DC between D and C.
7 This choice of a construction may seem as if it comes out of the blue, but in fact, it is similar to the way
we solved Fermats Problem the first time.

CHAPTER 8. INVERSION IN A CIRCLE

178

A
X
D

But if X is on DC, then the sum

CX + DX on the right of equation 8.6 is
constant, and the Ptolemys theorem extension tells us that the sum on the left
can be minimized by choosing X to lie
on the circle. Thus we have solved Fermats Problem.

Figure 8.14: Fermats Problem

Inversion/Ptolemyfermat.D [D]

8.10 Inversion to Concentric Circles

Given two circles in the plane that do not

intersect, find an inversion that maps them into a pair of concentric circles.
Concentric circles are perpendicular
to their diameters, so it seems clear that
K
M
the center of inversion we seek should
P
lie on the line connecting the centers of
the two original circles. That way, after
X B
O
A
the inversion, at least both the edges of
the images of those circles will remain
N
perpendicular to that line. But we would
also like to find a circle center such that
after the inversion a line perpendicular
to the line connecting the centers will
also be orthogonal to both the images of Figure 8.15: Inversion to Concentric Circles
Inversion/Concentric.T [M]
the circles.
From any point on the radical axis
(see Section 5.7) we can find a circle centered there that is also perpendicular to both
circles. Since the goal is symmetry, choose the circle whose center is on the intersection
of the radical axis and the line connecting the diameters of the two circles.
Figure 8.15 makes this clear. The original circles are centered at A and B. To find
the radical axis, connect the midpoints M and N of their common external tangents.
But M N intersects AB at O. Now find the circle centered at O that is orthogonal to
both circles. (To do this, find the point P where a line from O would be tangent to one
of the circles, and draw the circle centered at O and passing through P . Note that since
O is on the radical axis, the lengths of the tangent segments to both circles will be the
same, so a circle of that radius will be perpendicular to both.)
Let this circle intersect the line AB at X, and all we need to do is invert both circles
in an arbitrary circle K centered at X. Since AB passes through X and AB is orthogonal to both circles, the line AB will invert to itself and hence will remain perpendicular
to the images of both circles. Since the circle centered at O passes through X, under
inversion it will go to a line. That line will be perpendicular to AB, since the circle
that inverted to it was.

8.11. THE STEINER PORISM

179

But that line is the image of a circle that was also perpendicular to both the original
circles, so the line will be perpendicular to their images. Thus, after inversion, we know
of two perpendicular lines that are both perpendicular to the images of both circles, and
the only way this can occur is if the two circles are concentric.

8.11 The Steiner Porism

We now have all we need to prove a truly
amazing Euclidean geometry theorem.
Theorem 8.3 (Steiner Porism) Suppose we
are given two circles K0 and K1 where one
is completely contained within the other, and
which do not touch. Construct an arbitrary
circle between the two that is tangent to both,
and call it C0 . Now construct a circle C1 that
is tangent to K0 , K1 , and C0 . Next, continuing away from C0 construct a circle C2 tangent to K0 , K1 , and C1 . Continue in this way
as shown in figure 8.16.

K1

Cn

Cn-1

C0 K0
C1
C2

C3

Figure 8.16: The Steiner Porism

Inversion/Steiner.T [S]

It may happen that for some n, Cn is exactly tangent to C0 as occurs in the figure. If
that is true, then it does not matter where you
chose the initial circle C0 . The sequence of circles will have Cn tangent to C0 independent of the starting position. Furthermore, if for no n is Cn tangent to C0 then
modifying the starting point will not helpsuch a tangency will never occur 8 .
Based on the previous problem, this one is fairly simple. Find an inversion that
takes circles K0 and K1 into concentric circles. Then it is clear that the starting position
does not matterall the Ci will be exactly the same size, and they either meet with
tangency or they do not. If they do, then they can be inverted back through the circle
of inversion to place the image of C0 wherever is desired. If they do not, then no such
inversion will work.

8.11.1 Drawing the Steiner Porism

How can figure 8.16 be constructed using Geometer?
8 It may even be the case the tangent circles go around the loop more than one time before the n th one
is tangent to C0 . The theorem still holds, but from any starting point, a different set of tangent circles will
make the same number of loops before the nth one is tangent to the initial one.

CHAPTER 8. INVERSION IN A CIRCLE

180

Obviously, a set of equally-spaced

3
n
30
circles that exactly fill the ring between
A
two concentric circles was constructed,
M
B
and the result was inverted to obtain a
Steiner porism with a lopsided pair of enclosing rings. The inverted circles will
O
exactly fill the space between the lopsided rings.
It is not hard to work out the relative
sizes of a pair of concentric circles that
will allow for some fixed number n of circles to fit between them. In figure 8.17 we
see what we need to begin. In this case,
Figure 8.17: A Ball-Bearing Race
there are nine circles, but let us just call
Inversion/Steiner1.T [M]
that number n. The central angle between
any pair of circle centers is 360 /n, so
angle AOB = 360/n is a suitable vertex angle of the isosceles triangle in the figure.
If that figure, 4OM A is a right triangle where
M is the point of tangency of two adjacent circles. Suppose the radius of the inner circle is R1
C1
and
the radius of the small surrounding circles is
C2
C4
R2 (which will make the radius of the larger circle
C3
R1 + 2R2 ).
Then
 180 
R2
Figure 8.18: Four Circle Problem
sin
=
,
Inversion/Fourcircle.D [D]
n
(R1 + R2 )
and we can solve for R2 :

R1 sin( 180
n )
.
R2 =
180
(1 sin( n ))

Thus, if you can draw something like figure 8.17, it can be inverted through some
circle to produce the required general figure.
The only tricky part of the construction was the method by which the number of
circles in the loop could be altered by the Geometer program. As the point that controls
the number of inner circles is moved on the line, a ratio is calculated of its position
relative to the length of the entire segment. That value is truncated to obtain the number
of circles to draw. Then the largest number of circles is drawn, each offset from the next
by the same angle. If the actual number is smaller than the largest possible number,
extra circles are simply drawn on top of earlier ones so only the smaller number appear
to be drawn.

8.12 Four Circle Problem

8.12. FOUR CIRCLE PROBLEM

181

See figure 8.18. Given a line, construct a circle tangent to it centered at

C1 with radius r1 . Next, construct a
circle tangent to both the line and the
C1
B
C2
circle centered at C1 . The new circle
C4
has center C2 and radius r2 . As shown
C3
in the figure, construct the circle cenA
tered at C3 of radius r3 tangent to both
circles and to the line, as shown. Finally, the circle centered at C4 of raFigure 8.19: The Four Circle Problem
dius r4 = 1 cm is tangent to the first
Inversion/Fourcircle1.D [D]
three circles and lies inside as in the
figure. Find the perpendicular distance from C4 to the line in terms of r1 , r2 , r3 ,
and r4 = 1 cm.
Although at first it does not look like it, this is just a special case of the Steiner
porism, but in this case, the outer circle is the circle with an infinitely large radius
the straight line. Imagine what would happen if this figure were inverted in a random
circle that was not special in any way (none of the lines or circles pass though its
center). It would become the simplest example of a Steiner porism with an inner and
outer circle and three circles filling the ring between them.
If that is the case, it does not matter what r1 , r2 , and r3 arethe height above the
line will only depend on the radius r4 = 1 cm.
So we might as well choose a set of circles with convenient radii as in figure 8.19,
where r1 = r2 .
Using the pythagorean theorem on the right triangle 4C1 BC3 we have:
r12

r12 + (r1 r3 )2

r12

r32

2r1 r3 +
r12 4r1 r3

r1 (r1 4r3 )
r1

= (r1 + r3 )2
= r12 + 2r1 r3 + r32
= 0
= 0
= 4r3 .

Now use the pythagorean theorem again, but this time on 4C1 BC4 :
r12 + (r1 2r3 r4 )2
+ (r1 r1 /2 r4 )2

r12

r12 + (r1 /2 r4 )2
r12 + r12 /4 r1 r4 + r42

r12 /4 3r1 r4
r1 (r1 12r4 )

= (r1 + r4 )2
= (r1 + r4 )2
= (r1 + r4 )2
= r12 + 2r1 r4 + r42
= 0
= 0

r1 = 12r4 .

But r4 = 1 cm, so r1 = 12 cm and r3 = 3 cm. The height of C4 above the line is

7 cm.

CHAPTER 8. INVERSION IN A CIRCLE

182
The solution above seems pretty
good, but there is an even easier way.
See figure 8.20. We can invert to a situation where two of the circles become
straight lines, and the calculations become even easier. If the radius of the
circle centered at C4 is 1 and the unknown equal radii of the other two circles are both r, we can see that r satisfies:
2

C4
C3

A
C2

Figure 8.20: The Four Circle Problem

r + (r 1) = (r + 1) .

Inversion/Fourcircle2.D [D]

The solution is r = 4 and from the figure it is clear that the point C4 is 7 cm above the
lower line.
Note that this problem could also have been solved using the Descartes circle theorem presented in the form of a poem at the beginning of this chapter. In this case,
one of the circlesthe straight linehas infinite radius, so 1 divided by that radius
is zero.

8.13 The Arbelos of Pappus

Arbelos is the Greek word for a
O
shoemakers knife. In figure 8.21,
ignore everything except for the
three circles with diameters OA,
OB, and AB, where O, A, and
C-3
C3
B lie on the same line, and notice that the area inside the larger
circle and outside the two smaller
C-2
C2
circles is divided into two pieces
on the left and right. Either of
these shapes, which are basically a
A
C-1
C1
half-circle with two half-circles removed, look something like an arC0
belos.
From those original three circles, construct a series of circles
B
on both sides of the circle with diameter AB as shown in the figure.
Figure 8.21: The Arbelos of Pappus
Each circle is tangent to the circles
Inversion/Arbelos.T [M]
with diameters OA and OB and
also tangent to the previous circle
in the series. In the figure, the centers of those circles on the right are labeled C 1 , C2 ,
et cetera, and those on the left have centers C1 , C2 , . . . .

8.13. THE ARBELOS OF PAPPUS

183

We will show that the distance between C1 and C1 is twice the diameter of the
circle centered at C1 , that the distance between C2 and C2 is four times the diameter
of the circle centered at C2 , and in general, that the distance between Cn and Cn is
2n times the diameter of the circle centered at Cn . In the figure, this is illustrated for
C3 and C3 and the circles centered thereexactly 5 circles of the same diameter as
those circles can be placed on a straight line between them. There would be 1 between
the circles at C1 and C1 , 3 of them between the circles at C2 and C2 , et cetera.

The proof is not difficult, and since we have been looking at inversion and the
Steiner porism, it is clear that the situation here is very similar. If, for example, we
can find an inversion that leaves the circles centered at C3 and C3 fixed and at the
same time maps the circles with diameters OA and OB into parallel lines, we will be
done. The circles between those centered at C3 and C3 , namely those centered at
C2 , C1 , C0 , C1 , and C2 , under inversion will remain tangent to their neighboring
circles, and to the two parallel lines. Clearly, when we look at C4 and C4 , there will
be two more circles in the chain between them, so they will have two more circles on
the line between them.

C-3

C3

C-2

C2
R
C-1

C1

C0
B
Figure 8.22: Proof of the Arbelos Property
Inversion/Arbelos1.T [P]

It is not hard to find such an inversion. Clearly, it will have to send the circles with
diameters OA and OB to parallel lines, so the circle of inversion must be centered at
O. Figure 8.22 shows the inversion of the circles Ci in such a circle centered at O and
passing through a point R.
Since the circles Ci and Ci are symmetric relative to the line OB, as the radius
of the circle of inversion centered at O increases, it will expand through each pair of
opposite circles in exactly the same way. At some point as it expands through each
pair, it will be orthogonal to both and at that point, both of those circles will be inverted
into themselves.

184

CHAPTER 8. INVERSION IN A CIRCLE

This result was known to Pappus sixteen centuries ago, and he certainly did not have
the method of inversion available to him. To learn how Pappus might have done it, take
a look at How Did Pappus Do It?, a chapter by Leon Bankoff in [Klarner, 1981].

8.14 Another Arbelos Result

Beginning with the same arbelos (see figure 8.23), construct a perpendicular to the
diameter AB at C, and then construct the two circles (shown in the figure with centers
at D and E) tangent to that line and to the other pairs of circles. Show that the two
circles with centers at D and E have the same radius. They are called Archimedes
circles.
A little bit of fiddling around with a computer geometry program leads to an unusual
D
E
proof. If we draw the common tangent to the
T
circles centered at E and O, we notice that it
always seems to pass through the point A. We
A
Q C XO B
do not know if that is true, of course, but if it
is, we can just write down equations and solve
for the radius of the circle centered at E as follows:
Let R1 be the radius of the circle with diameter CB, R2 the radius of the circle with Figure 8.23: Another Arbelos Propdiameter AC, and r1 is the radius of the circle erty
centered at E.
Inversion/Arbelos2.T [M]
If the common tangent through T does
pass through A, since EO is perpendicular to it we have 4XEO 4T AO by AA,
since there is a common right angle and shared angle XOE, so:
XO
EO
=
OT
AO
R1 + r 1
R1 r 1
=
R1
2R2 + R1
(2R2 + R1 )(R1 r1 ) = R1 (R1 + r1 )
2R1 R2 r1 R1 2r1 R2 = r1 R1
2r1 (R1 + R2 ) = 2R1 R2
R1 R2
r1 =
R1 + R 2

This is a nice result, since it shows that the radius is a symmetric function of R 1
and R2 . Exactly the same calculations would give the radius of the circle centered at
D to be the same thing, since we just need to exchange R1 and R2 which leaves the
final result unchanged.
We still have not proved anything since we do not know that the tangent at T passes
through A. However, we will have a proof if we can show that the distance from E

8.15. THE PROBLEMS OF APOLLONIUS

185

to the large circle is also equal to r1 . To do that, we just need to show that QE =
R1 + R2 r1 . (The radius of the large circle is R1 + R2 , and if the circle centered at
E is tangent to it, then the radius of the large circle from Q through E had better cross
E a distance r1 from the outer circle.) Using the pythagorean theorem twice, we have:
EX 2 + XO2
EX 2 + QX 2

= EO2
= QE 2

(8.7)
(8.8)

We know that EO = R1 + r1 , XO = R1 r1 , and QX = R2 R1 + r1 . If we

just plug these values into equations 8.7 and 8.8 we obtain:
EX 2 = (R1 + r1 )2 (R1 r1 )2 = 4R1 r1
so

QE 2 = 4R1 r1 + (R2 R1 + r1 )2
We are done if QE = (R1 + R2 r1 ) which is true if:
(R1 + R2 r1 )2 = 4R1 r1 + (R2 R1 + r1 )2 ,

which we can check:

(R2 + (R1 r1 ))2 (R2 (R1 r1 ))2 = 4R1 r1
4R2 (R1 r1 ) = 4R1 r1
R1 R2
,
r1 =
R1 + R 2
which is true.

8.15 The Problems of Apollonius

Given any combination of three figures chosen from among points, lines, and circles,
find all the circles that pass through the selected points and are tangent to the selected
lines and circles. There are ten possibilities which we discussed in Section 3.6.5, labeled here as follows: (P P P ), (P P L), (P P C), (P LL), (P LC), (P CC), (LLL),
(LLC), (LCC) and (CCC). The (P LL) problem, for example requires the construction of all circles passing through a given point and tangent to two given lines.
The (P P P ) and (LLL) problems are solved in every high school geometry course.
In Sections 3.6.5 and 3.6.6, we solved (P P L) and (P LL), and in Section 8.6 we solved
(P CC). In Section 8.7 we even solved a special case (CCC).
Although all ten problems can be solved with classical techniques, the technique of
inversion simplifies many of the problems. In this section we will solve (and in some
cases, re-solve) various Apollonius problems using that technique.
In many cases, we can consider lines and circles to be basically the same, from
the point of view of inversion, so if you can solve, say, the (P CC) problem, you can
probably also solve the (P LL) and (P LC) problems, although you must be careful.

186

CHAPTER 8. INVERSION IN A CIRCLE

Here is a short discussion of all the cases. The one that is commonly known by
itself as Apollonius Problem is the hardest one: (CCC). It has approximately the
same difficulty as (LLC), or (LCC). In fact, if you have a rock-solid solution to the
three circles problem, you could solve the other two as follows:
First, do an inversion around an arbitrary point not on any of the lines or circles,
and convert your problem to (CCC). Solve that problem with your rock-solid method,
and then that solution can be re-inverted through the circle around the arbitrary point
that you initially selected. Notice that since your rock-solid method may itself involve inversion and re-inversion, you may have to do four total inversions to solve the
problem.
In what follows, we may ignore the degenerate cases, such as when the points lie
on the lines or circles, when the points are coincident, or when the three lines or circles
meet at a point. These cases are not hard, and you should be able to figure them out
without much difficulty.

8.15.1 Apollonius: (P P P ) and (LLL)

The (P P P ) and (LLL) problems are equivalent to finding the circumcircle and incenter, respectively, of a triangle.
The (P P P ) problem can be solved using inversioninvert in a circle centered at
one of the points, sending that point to infinity. If the line connecting the inversions of
the other two points is re-inverted in that circle, it will be the required circle. Do you
see why?

8.15.2 Apollonius: (P P L)
Do an inversion centered at one of the points. This will send that point to infinity, and
will convert the line into a circle. The inverted solutions must pass through the point
at infinity, and hence must be lines. They must pass through the inversion of the other
point and be tangent to the circle, so if we can find the tangents from the inverted point
to the circle, and re-invert that, we will have the solutions. There may be none if the
inverted point lies inside the inverted line (which is a circle).
It is not hard to consider the degenerate case where one of the points falls on a line.
Invert through the other point, and now there will be only a single solution, since the
inverted point will lie on the circle that is the inversion of the line, and there is hence
only a single tangent line.

8.15.3 Apollonius: (P P C)
This is exactly the same as the previous problem, except that the circle will be inverted
to become a circle instead of the line becoming a circle.

8.15. THE PROBLEMS OF APOLLONIUS

187

8.15.4 Apollonius: (P LL), (P LC) and (P CC)

The (P LL) problem was solved in Section 8.6, but it, and the other two, can be solved
via inversion. Send the point to infinity, and the lines and/or circles will become circles. Find the common exterior tangents, re-invert them, and obtain the solutions. The
special cases where a line or circle passes though the point are also easily solved.

8.15.5  Apollonius: (LLC), (LCC) and (CCC)

All three of these problems are equivalent in the sense that a single inversion can convert all of them to (CCC). If the problem initially has any lines, do an inversion
relative to a circle K whose center does not lie on any of the lines or circles. All the
lines and circles will be inverted to circles. We will demonstrate below how to solve
the three circles problem, and once you have done that for the inverted figures, re-invert
them in the circle K to obtain your final solution.
There are up to eight possible solutions, and figure 8.24 illustrates a situation where
all eight circles tangent to the three given circles exist.
The key idea is this: If the circle of smallest radius r is shrunk it to a point, and at
the same time r is either added or subtracted
from the radii of the larger circles, then solving
Apollonius problem for a point and two circles will yield a circle whose radius can be increased or decreased by r to yield a solution9 .

C1
R1
C3
R3

R2
C2

Figure 8.25 demonstrates the general idea.

The original circles for which the problem is
to be solved are centered at C1 , C2 , and C3 , Figure 8.24: The Problem of Apolloand they have radii C1 R1 , C2 R2 , and C3 R3 , nius
respectively. Assume that C1 R1 is the smallest
Inversion/Apollonius.T [M]
of the three radii. One solution can be obtained
by drawing a circle centered at C2 of radius
C2 R2 C1 R1 and by drawing a circle about C3 of radius C3 R3 + C1 R1 . Using
techniques we learned earlier in the chapter, we can find a circle that is tangent to those
new circles with the modified radii and passing through C1 as shown in the figure. If
the radius of that circle is then increased by C1 R1 , we have one of the eight possible
solutions.
This is trickier than it sounds. Remember that there are up to four solutions to the
(P CC) problem, possibly having tangencies on both sides of both circles. Only one of
these four circles can have its radius increased and still be tangent to the original circles.
In fact, if you play with the Geometer diagram that generates figure 8.24, you will
9 There are direct solutions for (LLC) and (LCC) as well that involve moving the line or lines parallel
to themselves by a distance equal to the diameter of the smallest circle, just as we expand and shrink the
diameters of the larger circles in the three-circle solution.

CHAPTER 8. INVERSION IN A CIRCLE

188

find that it works only for a limited range of values around the initial configuration
increase or decrease the radii too much and you will find that the solutions jump to
the other sides of circles, and on expansion or contraction of those circles, they are no
longer solutions to the problem.

R3+R1
R3

A different solution based on poles and polars appears in Section 10.6.

R1
C1

C3
R2
C2
R2-R1

8.16 Peaucelliers Linkage

For many years, it was unknown whether it

was possible to construct a mechanical linkage
that would turn perfect circular motion into
perfect linear motion. Peaucelliers Linkage
Figure 8.25: Addition and Subtrac(see figure 8.26) shows a linkage that achieves
tion of Radii
this conversion.
Inversion/Apollonius1.T [M]

In the linkage, point O is fixed on the circle, point A is constrained to move on the circle, and segments OC and OD are two bars of
the length l, while segments AC, CA0 , A0 D,
and DA are four bars of length r. The bars are
all hooked together with flexible joints at points O, A, C, A0 , and D. (The lines OA0
and CD in the figure are solely for the proofthey are not part of the linkage.)
We can show that the point A0 will
lie on a straight line if we can show that
OA
OA0 is constant. If that is the case,
A
E
A
then A and A0 are inverse points with reO
l
D
r
spect to a circle centered at O. As the point
A moves on a circle that passes through O,
its inverse, A0 must move along the inverse
of that circle, which is a straight line since
O lies on the circle upon which A is conFigure 8.26: Peaucelliers Linkage
strained to lie. (If O is not on the circle, A
Inversion/Peaucellier.T [P]
and A0 will still be inverse points relative
0
to a circle centered at O, but A will merely move on a different circle as A traces out
the first one.

To show this, construct the lines OA0 and CD. Since ACA0 D is a rhombus, CE
OA0 and E bisects AA0 . Thus we have
OA OA0 = (OE AE)(OE + EA0 ) = (OE AE)(OE + AE) = OE 2 AE 2 .
Using the pythagorean theorem on 4AEC and on 4OEC, we have
OE 2 + EC 2
AE 2 + EC 2

= OC 2 = l2
= AC 2 = r2 .

8.17. FEUERBACHS THEOREM

189

Subtracting, we obtain:
OE 2 AE 2 = l2 r2 = OA OA0 .
Since l2 r2 is constant, so is OA OA0 and we are done.

8.17  Feuerbachs Theorem

Feuerbachs theorem has a fairly difficult proof, both because there are a lot of
steps, and because a couple of those steps
depend on big theorems themselves.

Figure 8.27: Feuerbachs Theorem

Inversion/Feuer.T [M]

The theorem is not difficult to understand; figure 8.27 shows the nine-point
circle and the four circles that are tangent
to the three sides of the trianglethe incircle and the three excircles. Remember
that the nine-point circle passes through
the midpoints of the sides of the triangle.

Theorem 8.4 (Feuerbachs Theorem) Given an arbitrary 4ABC, the incircle and
the three excircles are all tangent to the nine-point circle. The point of tangency between the incircle and the nine-point circle is known as the Feuerbach Point.

8.17.1 Analysis of the Problem

If we can show that the nine-point circle is tangent to the incircle and an arbitrary one
of the excircles we are done, since all of the excircles are effectively equivalent.
Testing figure 8.27 yields zero relationships, interesting or otherwise, so we will
need to add a few things to it to get started. The nine-point circle passes through the
midpoints of the sides, so we might as well identify them, and it is probably useful to
indicate the points of tangency of the incircle and the excircle with the original triangle.
The result is illustrated in figure 8.28.
When Geometer is used to test this slightly more complex diagram, suddenly, there
are many relationships. Since 4A0 B 0 C 0 connecting the midpoints is similar to the
original triangle there will obviously be a lot of equal and supplementary angles, but
there is one other very interesting equality:
Equal length segments:
[D C] [E C]

Recalling that the subject of this chapter is inversion in a circle and that the incircle
and excircle are both tangent to AB at D and E, since C 0 is equidistant from D and
E, if we invert through a circle K centered at C 0 passing through D and E it will

CHAPTER 8. INVERSION IN A CIRCLE

190

A
D
B

C
E

Figure 8.28: Feuerbachs Theorem

Inversion/Feuerbach.T [M]

leave both the incircle and excircle fixed since both will be orthogonal to it. Since the
nine-point circle passes through C 0 , the nine-point circle will invert to a line through
C 0 . If, in fact, the nine-point circle is tangent to both the incircle and excircle, so will
be the line to which it inverts and that line thus must be a common internal tangent to
the incircle and excircle (AB is the other common internal tangent).
F
If we draw that circle of inversion, the inverted nine-point circle, and identify a
few important points we arrive at figure 8.29. In that figure (if the theorem is true),
GX is the inversion of the nine-point circle where X is the intersection of the common
internal tangents.

A
D
X
B C

A
C

E
B

Figure 8.29: Feuerbachs Theorem

Inversion/Feuerbach.T [M]

The easiest way to proceed is simply to draw the other external tangent and prove
that it is the inversion of the nine-point circle. To do that, we just need to show that a
pair of points on the nine-point circle invert to be on that line. Since inversion takes
points directly toward or directly away from the center C 0 , let B 00 and A00 be the intersections of the lines B 0 C 0 and A0 C 0 with the common tangent CX. If we can show
that B 0 C 0 C 0 B 00 = A0 C 0 A00 C 0 = C 0 D C 0 D we will be done.

8.17. FEUERBACHS THEOREM

191

Here we will just show that B 0 C 0 C 0 B 00 = C 0 D C 0 D since the proof for the other
pair is completely analogous.
If we test the new diagram in figure 8.29 there is a great deal of interesting information, a small amount of which is listed below which certainly shows us that we are
on the right track.
Supplementary angles:
(C A B) (C B A) (C G F) (C A B)
are supplementary to
(A C A) (F B C)
Points B and B are inverse relative to circle c4.
Points A and A are inverse relative to circle c4.

In what follows, we will use the standard notation that AB = c, BC = a and

CA = b. We will also make use of the relationships illustrated in figure A.19 in
Appendix A.
We can find the lengths of C 0 D = C 0 E, of C 0 B 00 and of C 0 B 0 in terms of s, a, b
and c and show that B 00 is the inversion of B 0 relative to the circle K centered at C 0 .

8.17.2 Proof of Feuerbachs Theorem

As stated above, we want to show that:
B 0 C 0 C 0 B 00 = C 0 D C 0 D

so we would like to find these lengths in terms of a, b and c.

From figure A.19 we know that AD = BE = s a, and since AC 0 = C 0 B = c/2,
0
C D = C 0 E = (ab)/2. We also know that B 0 C 0 = a/2, but obtaining an expression
for C 0 B 00 is a bit trickier.
By AA similarity, 4C 0 XB 00 4BXG, so:
BG
C 0 B 00
=
.
0
CX
BX

(8.9)

Since C 0 X = BX C 0 B = BX c/2, if we can find BX and BG in terms of a,

b and c, we will have all the relationships we need.
The point X is at the intersection of the common internal tangents, so X lies on the
angle bisector of BCA. From this, we know that
BX
AX
=
.
BC
AC
Since AX + BX = c and AC = b we can solve equation 8.10 for BX:
BX
c BX
=
CX
b
b BX = a(c BX)
BX(a + b) = ac
ac
.
BX =
a+b

(8.10)

CHAPTER 8. INVERSION IN A CIRCLE

192

To obtain a value for BG, notice from figure A.19 that BG = (s b) (s a) =

a b. If we put this together with equation 8.9, C 0 X = BX c/2 and the result above
that BX = ac/(a + b), we obtain:

C 0 B 00

=
=
=
=

BG C 0 X
BX
(a b)(BX c/2)
BX
(a b)c(a b)/(2(a + b))
ac/(a + b)
2
(a b)
.
2a

Now we can finish the proof:

B 0 C 0 C 0 B 00 =

a (a b)2
(a b)2

=
= C 0 D C 0 D.
2
2a
4

8.17.3 Feuerbach Consequence

Let H be the orthocenter of 4ABC. Then the nine-point circle of 4ABC is tangent to
all twelve of the incircles and excircles of the triangles 4ABH, 4BCH, and 4CAH.

Using Feuerbachs theorem, this is

pretty easy to showsee figure 8.30. The
nine-point circle of the original triangle is
the same as the nine-point circle of each
of the smaller triangles, since the ninepoint circle passes through the midpoints
of the segments connecting the orthocenter with the vertices. Since it is still a
nine-point circle, it is tangent to all the
circles named. (To avoid too much clutter, the figure includes only one typical
incircle and one typical excircle.)

B
H

A
Figure 8.30: Twelve Tangencies
Inversion/Ninefeuer.T [M]

Chapter

 Projective Geometry
This chapter will not supply many new techniques for using Geometer (although it
will provide a few). It does, however, provide some insight about how Geometer (and
probably most other computer geometry programs) do many of their internal calculations in terms of projective geometry as opposed to Euclidean geometry.
We will begin by proving the Euclidean version of a purely projective theorem. A
careful analysis of the theorem will then show why it is often useful to view a situation
as if it were in projective space rather than Euclidean space.

9.1 Desargues Theorem

Desargues theorem, like most of the others in this chapter, is actually a theorem from
projective geometry. We will begin by proving the Euclidean form of that theorem.
193

CHAPTER 9. PROJECTIVE GEOMETRY

194

Theorem 9.1 (Theorem of Desargues) Let 4ABC be an arbitrary triangle and P be

a point different from A, B and C. Choose points A0 , B 0 and C 0 on the lines P A, P B
and P C, respectively. Find the intersections of the corresponding sides of triangles
4ABC and 4A0 B 0 C 0 , if they exist. In other words, find the points at the intersections
of lines AB and A0 B 0 , of BC and B 0 C 0 and of CA and C 0 A0 . The points obtained
will lie on a straight line.
Of course one or more pairs of corresponding sides may be parallel so the points
of intersection do not exist.) If not all three intersections exist, the theorem is trivially
true, so let us assume that all three points of intersection exist. (See figure 9.1.

The proof invokes four applications of

Menelaus theorem. One of the problems
with Menelaus theorem is that it applies to
any triangle and any line that does not go
through one of the triangle vertices. In figure 9.1 there are a lot of triangles and a lot of
lines.

C
A C X
B
B

To use Menelaus theorem we need to discover length relationships between the points
X, Y and Z in the figure with some triangle.
In the diagram, these points lie on the lines
Figure 9.1: Desargues Theorem
that make up the sides of either 4ABC or
Projective/Desargues.T [P]
4A0 B 0 C 0 . Since the two triangles enter the
figure symmetrically, we can use either, so we
will arbitrarily choose to use 4ABC. The triangles are symmetric in this figure only
in the sense that if we exchange the labels A with A0 , B with B 0 and C with C 0 , the
theorem would be no different.

If we are to apply Menelaus theorem to 4ABC and what will be shown to be line
XY Z, we will need to find ratios that involve the lengths AX, CX, BY , CY , AZ
and BZ. Other interesting lines that are available to us include A0 B 0 Z, A0 C 0 X and
B 0 C 0 Y . We also need to make use of the fact that the point P is a center of projection
from 4ABC to 4A0 B 0 C 0 .
All of the conditions in the above paragraph are satisfied if we apply Menelaus
theorem to the following three pairs of triangles and sides: 4ACP and line A 0 C 0 X,
4CBP and line B 0 C 0 Y and finally 4P AB and line A0 B 0 Z. If we do so, we obtain
the following equations:
P A0 AX

AA0 CX
CY BB 0

BY P B 0
BZ P B 0

AZ BB 0

CC 0
P C0
P C0

CC 0
AA0

P A0

= 1

(9.1)

= 1

(9.2)

= 1.

(9.3)

9.2. PROJECTIVE GEOMETRY

195

If we multiply together the three equations 9.1 through 9.3, there is a great deal of
cancellation and we finally obtain:
AX CY BZ

CX BY AZ

1,

which, together with Menelaus theorem applied to 4ABC, implies that X, Y and Z
are collinear. In this book, of course, things work out smoothly, but when you try to
do this yourself the first time, it is almost certain that you will write down one of the
products of ratios upside-down and you might obtain the following, for example, in
place of equation 9.1:
AA0 CX P C 0

= 1.
P A0 AX CC 0
Obviously it is equivalent, but you will find that you sometimes need to invert the ratios
to make the terms cancel properly.
Note: There is an interesting arrangement of the code in the Geometer diagram for
this theorem. The text reuses the various expressions representing the results of different applications of Menelaus theorem and for this reason it is instructive to examine
the Geometer code in a text editor.

9.2 Projective Geometry

In this section we take a short but very interesting detour from Euclidean to projective
geometry that sheds light on the relationship between the two geometries. Later we
will see how projective geometry simplifies writing computer graphics code such as
that used in Geometer.
In the statement of Desargues theorem we had to be careful to consider situations
where some of the lines of interest were parallel since in those cases, certain of the
points of intersection did not exist. It is interesting to experiment with the Geometer
diagram to see exactly what happens when that occurs. If you move A 0 or B 0 to make
the line A0 B 0 approximately parallel to AB (keeping, for the moment, the intersections
of the pairs AC and A0 C 0 and of BC and B 0 C 0 ), it is certainly true that the point Z
moves farther and farther away, but when the lines are parallel and Z, in a sense, gets
to infinity, the infinity it reaches is in the directions of the lines AB, A 0 B 0 and XY .
In other words, the closer A0 B 0 is to being parallel to AB, the closer XY is to
being parallel to both of them. If we were to think of a point at infinity where the
parallel lines AB and A0 B 0 meet, it would have to be the same point at infinity where
AB and XY or A0 B 0 and XY would meet.
In Euclidean geometry, two lines can intersect at only one point and we would like
this to be true in our new geometry. If we add a point at infinity where parallel lines
meet, there would have to be a different point at infinity for every possible direction. If
non-parallel lines shared any point at infinity, they would meet at two pointsthe one
at infinity and their normal Euclidean intersection.

196

CHAPTER 9. PROJECTIVE GEOMETRY

Another nice feature of Euclidean geometry is that any two points lie on a unique
line. If we have added a point at infinity for every possible direction, on which line
do pairs of those points at infinity lie? The easy solution is simply to add a line at
infinity that includes all the points at infinity.
If we begin with plane Euclidean geometry and just add a point at infinity corresponding to every direction and then place all of those new points on a single line at
infinity, we have essentially constructed a perfectly good model for two-dimensional
projective geometry!
Note: There is only one point at infinity for any pair of parallel lines. In other
words, you arrive at the same point at infinity if you follow a line in either direction.
In a sense, the lines in projective geometry form loopsif you follow one to infinity
and then continue, you find yourself returning to the starting point from the opposite
direction.
The main difference between Euclidean and projective geometry is that in Euclidean geometry we have the following two axioms:
Every two points determine a unique line.
Every two lines determine a unique point, unless they are parallel.
In projective geometry, the corresponding axioms are completely symmetric:
Every two points determine a unique line.
Every two lines determine a unique point.
It is worthwhile to verify that these axioms hold in the model of projective geometry
that we just constructed by adding points at infinity to the Euclidean plane and the line
at infinity:
If the two points are Euclidean points, they lie on the normal Euclidean line that
passes through both of them. If one is a Euclidean point and the other is a point
at infinity, they lie on the Euclidean line passing through the Euclidean point
and going in the direction indicated by the point at infinity. If both points are at
infinity, they lie on the newly created line at infinity.
If two lines are intersecting Euclidean lines, they meet at the normal Euclidean
point. If they are parallel Euclidean lines, they meet at the appropriate point at
infinity. If one is the line at infinity, the lines meet at the point at infinity in the
direction of the Euclidean line. Since there is only one line at infinity, we need
not consider the situation where both lines are at infinity.
In fact, Desargues theorem is a pure projective geometry theorem. If we consider
the points and lines mentioned in the theorem to be projective points and lines, the
theorem holds in every case, whether the lines be parallel in the Euclidean sense or the
points be at infinity. (Note that in projective geometry there is no such thing as parallel
lines, since every pair of lines meet at a point.)

9.3. MONGES THEOREM

197

In the case we first considered in this section where the only problem was that lines
AB and A0 B 0 were parallel, their intersection point Z would still exist in projective
geometry and since XY in that case will be parallel to AB and A0 B 0 , XY will also
pass through the same point Z at infinity in that same direction.
If AB is parallel to A0 B 0 and BC is parallel to B 0 C 0 , then both Y and Z will be
(different) points at infinity. This means that X must also be a point at infinity, so AC
and A0 C 0 must also be parallel in this case. (This is not hard to prove.) Finally, and
this is left as an exercise to the reader, the point P could be at infinity, or so can some
of the vertices of these (projective) triangles. Think about a few of these and how each
situation would correspond to a Euclidean configuration.

9.3 Monges Theorem

At first glance, Monges theorem seems completely different from Desargues theorem
but in fact, they are virtually equivalent.
Theorem 9.2 (Monges Theorem) Given three circles in the plane, none of which intersect, none of which lie one within the other, and none of which have the same radius,
construct the common external tangents of each pair of circles. The three intersections
of those common pairs lie on a straight line. See figure 9.2.
The statement of the theorem can be
loosened somewhat. Pairs of circles can intersect and it is still true. If we make the
same restriction that we did in the statement
of Desargues theorem that if the external
tangents do not meet (in other words they
are parallel because the circles are the same
size), then the intersections lie on a straight
line (although in the case where there are circles with equal radii there can obviously be
only two or zero of them).

C
B
A

Figure 9.2: Monges Theorem

In fact, if two circles have equal radii,
Projective/Monge.T [D]
their external tangents meet at a point at infinity, and you will find that the other two
pairs of external tangents meet at points such that the line passing through them is parallel to the parallel pair of external tangents. If all three circles have the same radius,
the three sets of parallel lines meet at three different points at infinity, and those all lie
on the line at infinity, so there is a definite projective flavor to Monges theorem 1.
1 Monges theorem is not a theorem in projective geometry, since in projective geometry there is no such
thing as a circle. A circle can be projected to any conic section, and Monges theorem does not work if the
circles are replaced by ellipses or hyperbolas.

CHAPTER 9. PROJECTIVE GEOMETRY

198

C
Figure 9.3: Monges Theorem
Projective/MongeDes.T [M]

passing through that center of projection.

In figure 9.3 we see Monges

theorem again, but with some additional constructions. It is clear
that the line connecting the centers
of a pair of circles passes through
the intersection of their common
external tangents, and in fact, if
we think of that point of intersection as a point of projection, we
are really just projecting one circle onto another. If that is the case,
then any corresponding points on
the two circles will lie on a line

In the figure, a point drawn as a solid diamond is drawn in the corresponding position (at the same angle from the center) of each circle, and the lines connecting those
diamond points also pass through the common intersection of the external tangents and
the line of centers of the circles.
Now look at what is basically the same figure with some lines and circles erased
(see figure 9.4). In this figure it is obvious that the two triangles whose vertices are
all diamonds or all squares are projected from the same point at infinity (the lines
connecting the centers and corresponding points on the circles are parallel). Thus the
projective version of Desargues theorem (or a suitably extended Euclidean version)
shows that Monges theorem and Desargues theorem are equivalent.
One final interesting feature of
A
Monges theorem is that there is
a wonderful, obvious proof if we
first make the situation a bit more
complicated. Begin with the three
B
circles in a plane, but now consider them not to be circles, but
rather spheres that extend above
and below the plane. The exC
ternal tangents become cones that
touch each pair of spheres all the
Figure 9.4: Monges Theorem
way around, and it is clear that the
Projective/MongeDes1.T [M]
three-dimensional model of cones
and spheres, when cut by the plane
through the common centers of the spheres, gives the two-dimensional model with
which we began.
In three dimensions, consider two additional planes that touch all three spheres
above and below. Clearly each of these planes will be tangent to the cones and will
thus pass through all three tips. But two planes in three-dimensional space intersect in
a line, so the three tips of the cones must lie on that line.

9.4. THE THEOREM OF PAPPUS

199

9.4 The Theorem of Pappus

Let us take a look at another truly projective theoremthe theorem of Pappus.

X
A E

Z Y
C

Figure 9.5: The Theorem of Pappus

Projective/Pappus.T [M]

Theorem 9.3 (Theorem of Pappus) Let ABCDEF be a hexagon with points A, C,

and E on one line and points B, D, and F on a different one. Then the intersections of
the opposite edges of the hexagon (the intersection of AB with DE, of BC with EF ,
and of CD with F A) lie on a straight line. See figure 9.5.
The hexagon above is probably not what you normally think of as a hexagon, but
in projective geometry a hexagon is just a set of six distinct points connected by lines in
order. In such a general hexagon, the lines may or may not cross each other (and they
surely will cross each other in this example), but to satisfy the conditions of Pappus
theorem, they will clearly have to cross.
We can see that just as in Desargues theorem, the theorem of Pappus will usually
make sense in the Euclidean plane, but there will certainly be situations where one or
more pairs of lines are parallel so the points of intersection do not exist in the Euclidean
sense. The theorem as stated above, however, is always true in the projective sense,
even if, for example, the line containing the points A, C and E happens to be the line
at infinity.
Play with the Geometer diagram a bit to study its properties. Modify the order of
the hexagons vertices on the lines, and adjust the positions of the lines on which they
lie. Check to see that reasonable things happen (in the projective sense) when some
of the points or lines are the projective points or lines at infinity.
The theorem can, of course, be restated in a Euclidean form (which is the one we
will prove) as follows:
Theorem 9.4 (Theorem of Pappus in the Euclidean plane) If ABCDEF is a hexagon such that points A, C, and E lie one line and B, D, and F lie on a different line.
Then the points that lie at the intersections of AB with DE, of BC with EF and of
CD with F A, if they exist, all lie on a single line.

CHAPTER 9. PROJECTIVE GEOMETRY

200

Q
Z

X
A

Figure 9.6: The Theorem of Pappus

Projective/Pappus.T [M]

Notice that the Euclidean form could be extended to explain the situations that
could occur if various combinations of lines are parallel, but the complete statement
of the theorem in this form would probably cover an entire page. Such a complete
extension would have to include the following cases:
None of the corresponding pairs of sides are parallel.
One pair of corresponding sides is parallel.
If two pairs of corresponding sides are parallel, so is the third.
One of the points on a line is the point at infinity.
One point on each of the lines is a point at infinity.
One of the lines is the line at infinity.
The proof of the Euclidean version of Pappus theorem where nothing is parallel
basically involves an orgy of applications of Menelaus theorem (see figure 9.6). In
that figure, we need to show that the points X, Y , and Z are collinear, and we will do
so by showing that the conditions for Menelaus theorem hold for that line relative to
the triangle formed by the sides AB, CD, and EF , which meet at points P , Q, and R
in the figure. (We could equally well have used the triangle formed by the sides BC,
DE and F A.)
You have to fiddle around a bit to get this to work out nicely with the appropriate
things in the numerator and denominator, but the goal is to find the right number of
expressions of Menelaus theorem so that they can be multiplied or divided to give the
result. Looking ahead, we want to show that the three points X, Y , and Z lie on a line,
and Menelaus theorem says they will be on a line if and only if
P Y QX RZ

= 1.
RY P X QZ

(9.4)

Looking at 4P QR, we can find a bunch of lines that cut all three of its sides:
BY C, DXE, F ZA, DBF , and AEC. For each of these lines Menelaus theorem

9.5. ANOTHER VIEW OF PROJECTIVE GEOMETRY

201

will give a product of ratios that is equal to one. In the order above, those products of
ratios are:
QB RC

P B QC
QX RD

P X QD
QA RZ

P A QZ
P B QD

QB RD
P A QC

QA RC

PY
RY
PE
RE
PF
RF
RF
PF
RE
PE

= 1
= 1
= 1
= 1
= 1.

If we multiply all five equations above, everything cancels except for the desired
result: equation 9.4. Remember that when you are looking for a proof of your own,
you probably will not come up with exactly these five equationssome of yours may
be inverted and you will find you have to multiply and divide your equations, but the
net result will be the same.
Of course this is only a proof of the Euclidean version of the theorem, but the
theorem does hold in the general projective sense.

9.5 Another View of Projective Geometry

Imagine that you are looking out a glass window at the three-dimensional world outside, and then, without moving your head at all, you take a paintbrush and at every
point on the window, you paint on the glass exactly the color your eye sees there. If
your paint colors were accurate enough, if you were a good enough technician, and if
you were unable to move your head, you would be unable to tell if you were looking
at the scene outside the window or at the painting you had just made on the window
glass.
The image on the glass is a projection of the objects in the three-dimensional world
on the plane of the window, and if you wish to study this sort of operation, projective
geometry is just the thing. Three dimensional computer graphics software would not
exist without it.
Although projective geometry exists in any number of dimensions (in the example
above we were projecting the three-dimensional world to a two-dimensional painting),
in this chapter, unless we state otherwise, the words projective geometry will refer
to the two-dimensional version (two-dimensions projected onto two dimensions). We
can learn quite a bit by looking at one-dimensional projective geometry, and much of
that can be applied to higher-dimensional projective spaces as well. In fact, there are
even finite projective geometries where the entire geometry consists of a finite number
of lines and planes and theorems like those of Desargues and Pappus continue to be
true in those finite geometries as well. See Section 9.11.

CHAPTER 9. PROJECTIVE GEOMETRY

202

Go back to the painting example, but imagine that you are not looking at the world
through the window, but at a geometric diagram drawn on a flat plane. You can do
the same trick with the paintbrush and draw the image of the two-dimensional diagram
on the two-dimensional window, and this is a perfectly good projection too. As you
move your eye around behind the window, or as you move or tilt the glass window, the
painting you would obtain on the window will clearly change.

9.5.1 Geometric Invariants

One way to think about Euclidean geometry is that it is the study of properties of figures
that do not change if you move them around or rotate them (or even if you reflect them
in a mirror). For example, if two triangles are congruent, they will continue to be
congruent if you move them around without distorting them. In the same way, the
angles between the lines in a figure remain unchanged if you move the figure rigidly.
The areas of geometric objects are similarly unchanged if you move them elsewhere.
Other properties are also preserved, and in a sense they are almost too obvious to
mention. For example, if you rigidly move a line or a circle, it remains a line or a circle.
If you rigidly move two intersecting lines as a unit, they continue to intersect. There
are lots of other properties that are preserved under rigid motions.
Of course there are properties that are not preserved under rigid motionthe distance to a fixed point, for example, or which of two objects is to the right or left of the
other (since you can rotate the figure to reverse this).
What if the operations allowed include projections in addition to mirrorings, translations, and rotations? Now what properties are preserved?
Certainly not things like lengths, areas or angles. Projections of circles are very
rarely circles. Projections of parallel lines are not parallel.2 Even things like betweenness drop by the waysideif point B is between points A and C on a line, the projection of B may not be between the projections of A and C. They will, however, remain
on the same projected line.
But plenty of things are preserved (although we do need to make some very slight
changes to what we mean by points and lines):
If two lines intersect at a point, the projections of those lines intersect at the
projection of the point.
If two points lie on a line, the projections of those points lie on the projection of
the line.
Although circles are not preserved, conic sections are. In other words, if you
project any conic section (a circle, ellipse, parabola, or hyperbola), its projection
will also be a conic section, although any one of the types might be converted to
any of the other types.
2 In

fact, in true projective geometry, there really is no such thing as a pair of parallel lines.

9.6. PROJECTIVE DUALITY

203

Although betweenness is not preserved, something similar isseparation. If A,

B, C, and D lie on a line, and A and C are separated by B and D, their
projections satisfy the same separation condition. By A and C are separated by
B and D we mean, roughly, one of B or D is between A and C and the other
is not. Another way to think of it is to imagine that the line is bent into a circle,
then the points A and C determine two arcs, and if B is on one of those, D is on
the other.
Actually, many different geometries besides Euclidean and projective can be classified by the properties that are preserved under different types of transformation. Affine
geometry, for example, studies the properties that are preserved under transformations
that preserve parallel lines. In this book we will stick to Euclidean and projective geometry, and a bit of inversive geometry in the previous chapter.

9.6 Projective Duality

In Section 9.2 we stated a very interesting property of projective geometry theorems
the axioms are stated in such a way that points and lines are treated completely equivalently, and therefore if you take any projective theorem and swap the occurrences of
the words point and line you will get another theorem that is true.
So what is the dual of Pappus theorem? Try and figure it out before you read
the following description. Even better, try to make a computer geometry drawing that
illustrates it and play with it a bit.

cd
bc

ab

c
f

fa
a

de
Y

ef
b

Figure 9.7: Dual of the Theorem of Pappus

Projective/Pappusdual.T [M]

Theorem 9.5 (Dual of the Theorem of Pappus) Let lines a, c, and e pass through a
common point, and let lines b, d, and f all pass through another point. Next construct
three lines: the line connecting the intersection of lines a and b with the intersection of
lines d and e, the line connecting the intersection of the lines b and c with the lines e
and f , and the line connecting the intersection of the lines c and d with the intersection
of lines f and a. All three of those lines meet at a point.

CHAPTER 9. PROJECTIVE GEOMETRY

204

In figure 9.7, most of the important intersections are labeled according to the pairs
of lines that define them. The lines a, c, and e pass through point X and lines b, d, and f
pass through point Y . The intersection of lines a and b is labeled ab; the intersection of
lines b and c is labeled bc, and so on. As you can see in the figure, the lines connecting
appropriate pairs of intersections all meet at a point.
Since the projective axioms make no distinction between points and lines, a purely
projective proof of Pappus theorem also implies the truth of this dual theorem. If
this were only a Euclidean theorem, the proof could be carried out in the same way,
but using Cevas theorem a bunch of times. Cevas theorem is much like a dual of
Menelaus theorem, but remember that since Cevas theorem (and Menelaus theorem)
both involve lengths, they can not possibly be projective theorems.

9.7 Pascals Theorem

C

F
Z
B

K
X I

D
Figure 9.8: The Theorem of Pascal
Projective/Pascal.T [M]

Theorem 9.6 (Pascals Theorem) Given a hexagon all of whose vertices lie on a circle then the intersections of the opposite sides lie on a straight line.
In other words, if the points A, B, C, D, E, and F lie on a circle, and I is the intersection of AB and DE, J is the intersection of BC and EF , and K is the intersection
of CD and F A, then I, J, and K lie on a straight line.
As you can see from the figure, hexagon in this case simply means any six distinct
pointsthe lines connecting them can cross or not. (The reason the lines cross in the
figure is so that the intersections will be relatively close to the circle, but this is certainly
not requiredplay with a computer geometry diagram to convince yourself that the
ordering of A, . . . , F on the circle makes no difference.
Also note that the theorem as stated above is not true. There is no reason that the
opposite pairs of sides cannot be parallel. In the same way that the Euclidean version
of the theorem of Pappus can be written, so can Pascals theorem. It is not hard to do,

9.7. PASCALS THEOREM

205

and you should think of exactly how to state it. The statement begins with If none of
the opposite sides are parallel, and then has the statement above. Then work it out if
the statement begins, If one pair of opposite sides is parallel, and so on.
Finally, as stated above, the theorem is not really a theorem in projective geometry
either, since there is no such thing as a circle in projective geometry. We will talk about
that later. (In fact, it will turn out that the theorem of Pappus is just a special case of
the full-blown projective version of Pascals theorem that is true for hexagons on any
conic section.)
As in the case for Desargues theorem, we will just give the proof in the hard
casewhere all pairs of opposite sides are not parallel. If one or three pairs of sides
are parallel, things are much easier, and you can solve them yourself for extra credit,
if you wish.
The proof is extremely similar to the proof of Pappus theorem. The only thing we
have got to work with is some triangles and some transversals, and the fact that the
points lie on a circle, so after you come up with the key observation, the rest of the
proof is pretty mechanical.
Exactly as we did for Pappus theorem, consider 4XY Z that is made up of sides
AB, CD, and EF of the hexagon. In figure 9.8, X is the intersection of AB and CD,
et cetera.
It seems like for these projective-like theorems, the best way to prove the points I,
J, and K lie on a line is to use Menelaus theorem. Here is what we would like to get
(relative to 4XY Z):
ZI XK Y J

= 1.
(9.5)
IZ KY JZ
Relative to 4XY Z, we can also apply Menelaus theorem for three different
transversal lines: AKF , BJC, and DIE, yielding:
ZA XK Y F

AX KY F Z
ZB XC Y J

BX CY JZ
ZI XD Y E

IX DY EZ

= 1
= 1
= 1.

but we cannot get the two additional equations that we got for Pappus theorem since
the points do not lie on two lines, but rather on a circle.
We can multiply the three equations above (and rearrange it a bit) to get the following:
(ZI XK Y J) (ZA ZB) (XC XD) (Y F Y E)

= 1.
(9.6)
(IX KY JZ) (F Z EZ) (BX AX) (DY CY )
The grouping shows that we have exactly what we want on the left-most part of the
equation, but there are a bunch of other factors that had better turn out to be one.

CHAPTER 9. PROJECTIVE GEOMETRY

206

Luckily, the six points lie on a circle, so by looking at pairs of chords that intersect
at X, Y , and Z we get the following ratios which are exactly what we need to cancel
off all the unwanted pairs:
ZA ZB
XC XD
YF YE

= F Z EZ
= BX AX

= DY CY.

These, combined with equation 9.6, yield equation 9.5 and we are done.

9.7.1 Projective Version of Pascals Theorem

As we stated above, there is no such thing as a circle in projective geometry. Under
projections, circles are just one form of a conic section, and conic sections include
ellipses, hyperbolas, parabolas, and even degenerate forms, such as a pair of crossing
lines. But lines, under projection, remain lines.
Pascals theorem holds if the six points lie on any conic section. (If the conic
section happens to be a pair of crossing lines, Pascals theorem becomes the theorem
of Pappus.)

D
Z

F
B

X
A
Y

Figure 9.9: The Theorem of Pascal on an Ellipse

Projective/Pascalconic.T [M]

We will not prove the fully projective version here, but figure 9.9 shows an example
where the conic section happens to be an ellipse and the vertices of the hexagon are
in fairly general positions. (The unnamed points in the figure control the shape of the
conic section.) The Geometer diagram can be modified to illustrate Pascals theorem
on other conic sections.
Figure 9.10 illustrates a proof of a special case of Pascals theorem for a particular
hyperbola (y = 1/x), a particular point (the point at infinity in the direction of the
y-axis) and a particular line (the x-axis). In the figure, the point A is at infinity, and the
hexagon is ABCDEF . Lines AB and DE intersect at point , lines BC and EF at
, and lines CD and F A at , where , , and lie on the x-axis.

9.7. PASCALS THEOREM

207

A
E

D
B
C

Figure 9.10: The Theorem of Pascal on y = 1/x

Projective/Hyperbola.T [M]

As you can see, the positions of points D and E completely determine the location
of point B, and C and D determine the location of F , so since A is fixed at infinity in
the y-axis direction only points C, D and E can be moved in the Geometer diagram.
With this setup it is amazingly easy to show that lines BC and EF meet at a point
that is on the x-axis using just a little bit of analytic geometry.
For any two points P1 = (x1 , 1/x1 ) and P2 = (x2 , 1/x2 ) on the hyperbola, we
will find where the line P1 P2 intersects the x-axis. The equation of the line through P1
and P2 is given by:
!
1
1
1
x2 x1
y
=
(x x1 ).
x1
x2 x 1

Set y = 0 to find the intersection of the line with the x-axis, and we obtain:
 1 
1
=
(x x1 ),

x1
x1 x2
or
x = x1 + x2 .

In other words, if P1 has x-coordinate x1 and P2 has x-coordinate x2 , then the line
P1 P2 intersects the x-axis at the point having the sum of those coordinates.
Looking back at figure 9.10, let the x coordinates of the points B, C, D, E, and F
be b, c, d, e, and f , respectively. We know that e + d = b and d + c = f . To show that
lies on the x-axis, we need to show that the lines BC and EF intersect the x-axis at
the same x-coordinate. For BC, that point is at b + c, and for EF , it is at e + f . But
since e + d = b and d + c = f it is clear that b + c = e + f and we are done.

9.7.2 Brianchons TheoremPascals Dual

Since Pascals theorem is a pure projective theorem, it must have a dual. But what is
the dual of a conic section?

CHAPTER 9. PROJECTIVE GEOMETRY

208

First, we will show a Euclidean version of Brianchons theorem, then we will talk
about the dual of a conic section, and then we will show an example of Brianchons
theorem on a conic section.
de
d
cd

e
ef

bc
b

ab

fa

Figure 9.11: Brianchons Theorem on a Circle

Projective/Brianchon.T [M]

Figure 9.11 illustrates Brianchons theorem on a circle. a, b, . . . , f are six lines

tangent to a circle. The point labeled ab is where lines a and b intersect. The one
labeled bc is where lines b and c intersect, and so on.
If we connect opposite pairs of intersections (ab with de, et cetera), we find that
those three lines meet at a point.
The Brianchon projective dual version of Pascals theorem looks the sametake
six lines tangent to a conic, find the intersections of the adjacent ones, and connect
opposite pairs of those with lines, and those lines will all meet at a point. This is true,
but notice that it is not exactly stated as a dual theorem should be. Pascals theorem
begins Take six points of a conic section. . . . Brianchons theorem should then begin
Take six lines of a conic section. . . ? What is a line of a conic section?

Figure 9.12: A Conic of Lines

Projective/Coniclines.D [D]

The answer is that in projective geometry, there can be conic sections composed of
points and conic sections composed of lines. The ones composed of lines look to us
like the collection of all the lines that are tangent to the conic. Figure 9.12 shows a few
of the lines that compose an ellipse-like conic section. It all the lines were shown, of

9.7. PASCALS THEOREM

209

course, the only white portion of the figure would be the interior of the ellipse. Some
even better illustrations of this appear in figure 10.10.

e
b
de
ef

bc
cd

fa

c
f
Figure 9.13: Brianchons Theorem on a Hyperbola
Projective/Brianconic.D [D]

Finally, figure 9.13 illustrates Brianchons theorem on a conic section (a hyperbola,

in this case).

9.7.3  Degenerate Hexagons

Q
P

A
D

C
O

Figure 9.14: Pascals Theorem on a Degenerate Hexagon

Projective/Degenpascal.T [M]

There are a few interesting facts we can derive by applying Pascals and Brianchons theorems to degenerate hexagons. We will look at two examples. See figure 9.14. To do the proof, we actually use a bit of the theory of limits, which is why
there is a  in front of this section. It is a very straightforward use, however.
Let ABCD be a cyclic quadrilateral with no pair of parallel sides. Then the points
determined by the intersections of the lines AB and CD, of AD and BC, and of the

CHAPTER 9. PROJECTIVE GEOMETRY

210

tangent lines to the circle at points B and D (or of the tangents at A and C) all lie on
the same line.
Consider the hexagon ABBCDD, or, if you like, consider a hexagon ABB 0 CDD0
where B and B 0 are very close together and D and D 0 are similarly close. Then Pascals
theorem states that the intersections of AB and CD, of B 0 C and D0 A, and of BB 0 and
CC 0 all lie on a line. As B approaches B 0 and D approaches D 0 , clearly BB 0 and
DD0 approach the tangent lines, and B 0 C and D0 A approach BC and DA.
By relabeling, it is similarly easy to show that the tangents at A and C likewise
meet on the same line.

W
D

Z
C

Figure 9.15: Brianchons Theorem on a Degenerate Hexagon

Projective/Degenbrian.T [M]

As a second example, consider figure 9.15. Given any quadrilateral ABCD circumscribed about a circle, the diagonals AC and BD, and the lines connecting the
opposite points of tangency XZ and W Y all meet at a point.
Consider the degenerate hexagons AXBCZD and BY CDW A. Apply Brianchons theorem to both, and conclude that all four lines meet at a point.

9.8 Homogeneous (Projective) Coordinates

Just as in Euclidean space you can assign coordinates to your points so that you can
do mechanical calculations with them, the same thing can be done in projective space,
and the coordinates in this case are called homogeneous coordinates.
We will consider two dimensions here, but just by adding or deleting coordinates
the situation in more or fewer dimensions is very similar.
If you want to start with the Euclidean plane and add the points at infinity and the
line at infinity and to have a uniform method of dealing with them in homogeneous
coordinates, here is how to proceed.
Every point has three coordinates: (x, y, w) and the only constraint is that at least
one of the three has to be non-zero.3
3 In three dimensions, this becomes: Every point has four coordinates: (x, y, z, w), and the only constraint is that at least one of the four has to be non-zero.

9.8. HOMOGENEOUS (PROJECTIVE) COORDINATES

211

To convert a standard Euclidean point (not a point at infinity) to homogeneous

coordinates, take the two Cartesian coordinates and add a third coordinate of 1.
In other words, the point 3 units to the right and 7 units up from the origin which
was labeled (3, 7) in cartesian coordinates becomes (3, 7, 1) in homogeneous
coordinates.
Every point has many different representations in homogeneous coordinates. If
6= 0 is any real number, then (x, y, w) and (x, y, w) represent exactly the
same point. This is very much like what you do with fractions: 1/3 = 2/6 =
30/90.
Remember that every point at infinity can be thought of as a direction. From
the origin, pick a point (x, y) on the Euclidean plane, so that the line from the
origin to (x, y) goes in the direction you want. Then the point with homogeneous
coordinates (x, y, 0) represents that point at infinity. So for example, the point
at infinity whose direction is the bisector of the angle between the positive x and
y axes has homogeneous coordinates (1, 1, 0) (or (2, 2, 0), or (, , 0)all are
just multiples of each other).
Lines will also have three coordinates, and to differentiate them from points here,
we will enclose them in brackets, like this: [a, b, c]. Like points, any set of three
numbers represents a valid line as long as at least one of the three is non-zero.
Similarly, if 6= 0 is any real number, the lines [a, b, c] and [a, b, c] represent
the same line.
If the Euclidean line has the equation ax + by + c = 0, it will be represented by
the homogeneous coordinates [a, b, c]. In this case, it is obvious why [a, b, c] and
[a, b, c] represent the same lineone equation for the line is just a multiple
of the other.
The line at infinity can be represented using any
of the following sets of homogeneous coordinates [0, 0, 1], [0, 0, 2], or [0, 0, 17]they are all the same. This is
clearly not a Euclidean line since it would represent the equation 0x+0y+1 = 0,
or 1 = 0.
A point (x, y, w) lies on the line [a, b, c] if and only if ax + by + cw = 0. For
our standard translation of Euclidean points to homogeneous coordinates this
makes perfect sense. w = 1 and the x and y are the standard coordinates, so the
equation becomes ax+by+c1 = 0. Clearly only the points at infinity will lie on
the line [0, 0, 1]the two zeroes in the lines coordinates will eliminate anything
the point might have in its first two slots, and if the point is not at infinity, it will
have a non-zero number as its third coordinate making ax + by + cw 6= 0.
If (x0 , y0 , w0 ) and (x1 , y1 , w1 ) are two distinct points, then the line connecting
them is given by: [y0 w1 w0 y1 , w0 x1 x0 w1 , x0 y1 y0 x1 ]. This is easy to
checkjust make sure that both of the points lie on the line.4
4 Although this is not what is going on, if you know something about the three-dimensional vector crossproduct, this equation is easy to remember, since the line connecting the points A and B is given by A B,

212

CHAPTER 9. PROJECTIVE GEOMETRY

If [a0 , b0 , c0 ] and [a1 , b1 , c1 ] are two lines, then the point where they meet is
given by: (b0 c1 b1 c0 , c0 a2 a0 c2 , a0 b1 a1 b0 ). Notice (as must be the case)
that this equation and the previous are completely symmetric.
 If A is any 3-by-3 non-singular matrix, and if v is any point in homogeneous
coordinates, then vA represents a projective transformation (where v is considered to be a row vector). In other words, if you multiply every point in your
space by the same matrix A, you will get another space where all the projective
properties are preserved. If you transform all the points as above, you need to
transform the lines l by considering them to be column matrices and calculating
A1 l. This clearly works, since if a point v is on a line l, then (in matrix notation) v l = v(AA1 )l = (vA)(A1 l). If v l = 0 (the original point was on the
line), then the transformed point will be also, and conversely.

9.9 Higher Dimensional Projective Geometry

Like Euclidean geometry, we can consider projective geometry in 1, 2, 3 (and even
more) dimensions. To improve your intuition about geometry of any sort in higher dimensions, it is very useful to understand thoroughly what goes on in lower dimensional
situations. Unless you understand rotation in the two-dimensional plane, you will be
totally mystified if you try to understand it in three dimensions.
Many people are interested in the fourth dimension, but being fundamentally
three-dimensional creatures5 it is difficult to imagine what four dimensions might be
like.
Perhaps the best way to get a feeling for four dimensions is to look carefully at the
situation in all the dimensions available to you and perhaps you can see patterns that
might continue into the fourth dimension. Without too much trouble, you can usually
investigate the situation for many kinds of geometry (Euclidean, projective, and others)
in one, two, and three dimensions.
An example of this sort of approach to looking at four-dimensional Euclidean geometry can be found in Edwin Abbotts wonderful little book Flatland: A Romance
of Many Dimensions (see [Abbott, 1952]). Abbotts book was originally written in
the late nineteenth century, and has been updated by a book called Sphereland (see
[Burger, 1965]) that extends Flatlands ideas to Einsteins curved and expanding spacetime. Both books are worth reading. For those interested, a more advanced, but highly
visually-oriented book is called The Shape of Space (see [Weeks, 1985]).
Even three-dimensional projective geometry is a bit tricky to visualize (just ask
anyone who has tried to write code for a three-dimensional computer graphics system),
so we will begin our investigation by looking at the one-dimensional case in detail.
where is the three-dimensional vector cross product. Of course the same thing is true in the dual case
the cross product of two lines gives the point of intersection.
5 At least we are three-dimensional in spacetreating time as the fourth dimension and going into Einsteins theory of relativity is a bit beyond the scope of this book. Or if the string-theory guys are right, who
knows what our true dimension is.

9.9. HIGHER DIMENSIONAL PROJECTIVE GEOMETRY

213

Since we have already looked at the situation in two dimensions, we will then be in a
good position to talk about what goes on in three (and more) dimensions.

9.9.1 Basic 1-Dimensional Projections

Two-dimensional projective geometry
concerns projecting planes onto planes,
so one-dimensional projective geomeA
try concerns projecting lines onto lines.
A
P
B
This is very nice to work with, since
B
C
we can draw the entire situation on a
C
D
planesimply draw the two lines, the
D
E
point of projection, and a set of projecF E
tions, as in figure 9.16. In this figure,
P is the center of projection, and the
F
G
two one-dimensional lines that are proG
jected are AG and A0 G0 . This particular projection takes A to A0 , B to B 0 ,
and so on. Exactly the same illustration Figure 9.16: One-Dimensional Projections
Projective/Onedim.D [D]
can be interpreted as a projection from
0
the second line to the first, taking A to
A, B 0 to B, et cetera. In this particular case, the point P is in the Euclidean plane,
but the projection could be from a point at infinity, so all the projection lines would be
parallel, coming from the same direction.
Note that the projection maps all the points on one linenot just the half-dozen
that are shown in the figure. Every line through P projects a point on one line to a
point on the other. If we happen to choose a line through P that is parallel to one or
both of the lines, it maps the points at infinity appropriately.

8
7
6
5
4
3
2
1
0

8
7
6
5
4
3
2
1
0

Figure 9.17: A Scaling Projection

Projective/Scale.D [D]

In general, the two lines can be anywhere

in the plane, at any orientation. P can also be
anywhere as long as it is not on one of the lines
(otherwise a degenerate projection will occur
that maps all the points to a single point), and
in fact, P can even be interpreted as a point
at infinity so that the lines of projection are
parallel to each other.
Since our world is fairly nearly a Euclidean one (at least locally), it will be convenient to apply a coordinate system to the lines
to be able to describe how the points on the
lines are mapped by various sorts of projections.
Before we begin, however, we need to
look at one other thingare we going to need

CHAPTER 9. PROJECTIVE GEOMETRY

214

any points at infinity or lines at infinity? The answer is yes, but it is not so complicated as in the two-dimensional case. We need only add a single point at infinity to do
the job. You can think of the one-dimensional projective line as a loopif you go out
forever in the positive direction, you come to the point at infinity, and if you continue in
that direction, you will find yourself coming in toward 0 through the negative numbers.
Since we do not have a whole bunch of points at infinity corresponding to different
directions, but rather a single point, we will just call it here.
No line at infinity is required in 1dimensional projective geomety.

8
8
In figure 9.17 we see two lines, each of
7
7
which is fitted with the usual Euclidean coor6
6
dinate systemevenly spaced points. These
5
5
two happen to be parallel to each other, both
4
4
vertical, and aligned so that the two zero3
3
points are on the same horizontal line. In this
2
2
1
case, the center of projection, P , is on that
1
0
same horizontal line of the zeroes, such that
0
its distance to the first line is the same as the
distance from the first line to the second. It
should be clear that this particular projection
Figure 9.18: A Translating Projection
maps every point to a new point with double
Projective/Translate.D [D]
the coordinate of the original. (At least that
is if you project from the left line onto the
rightif you think of the projection as mapping the right line onto the left, then this
projection cuts every coordinate value in half.)
In mathematical terms, the particular projection in figure 9.17 could be denoted by
f (x) = 2xevery value is mapped to twice its original value by the projection.

8
7
6
5
4
3
2
0 1 2 3 4 5 6 7 8
P 0

It is not hard to convince yourself that

by moving P back and forth along the line,
the projection can represent any multiple you
want. In general, with a setup like this, we can
get anything of the form f (x) = x, where
> 0.
How about negative ? Well, if the second line were upside-down so that the negative coordinates were heading toward the top
of the page, we would have projections for
f (x) = x, where > 0.

What other operations could be obtained?

Is there a projection that just adds a constant?
Can we make a projection function that does
Projective/Invert.D [D]
this: f (x) = x + 3? Yes. See figure 9.18. The
two vertical lines are arranged the same way
as previously, but this time we just use a parallel projection that moves each point up
Figure 9.19: The 1/x Projection

9.9. HIGHER DIMENSIONAL PROJECTIVE GEOMETRY

215

three units. It should be easy to see that by changing the slope of the parallel projection,
an arbitrary amount can be added or subtracted, so in general, it is possible to have a
projection described mathematically as f (x) = x + , for any .
Let us consider one more example.
See figure 9.19. In this case, the two
lines are perpendicular, but they cross at
their 1 coordinate, and the center of projection, P , is where the horizontal and vertical lines through the zeroes would meet.
Again, without too much difficulty, you
can show that f (1) = 1, f (2) = 1/2,
f (3) = 1/3, f (1/4) = 4, et cetera. In
fact, this projection represents f (x) =
1/x. What happens when x = 0? Just
look at the diagramwhere would it be
mapped? The projection line is parallel to
the line it needs to hit, so the intersection
must be at the point at infinity, so effectively we have: f (0) = and f () = 0.

5
3

4
4

53

6
2

7
1

Figure 9.20: Three Point Projections

Projective/Threepoint.D [D]

9.9.2 Projection from 3 Points

How much freedom do we have in designing one-dimensional projections? The
answer is that you can take any three distinct points (including possibly the point at
infinity) and specify the three targets of those three points (which also must be distinct, but may include the point at infinity). A projection that achieves this is possible,
but once that much is specified, the projections of all the other points are completely
determined.
Figure 9.20 shows generally how to do it (the strategy has to be changed somewhat
for points at infinity, but not much). In this particular example, suppose we are seeking
a projection that makes the following mappings:
f (3) = 5 f (1) = 8 f (8) = 4.
Move the lines around so that the 3 on the first line lies exactly on top of the 5 on
the other. Now no matter where the center of projection is located, the first condition
in the list above will be satisfied.
Next, draw the lines connecting 1 on the first line with 8 on the second, and 8 on
the first line with 4 on the second. Let those lines intersect at P (and if those lines
are parallel, let P be the point at infinity). If P is the center of projection, it clearly
satisfies all the conditions of the problem. Furthermore, once we know where P and
the two lines are, the projection is completely determined.
Obviously, this does not prove that with another selection of orientations of the
lines that a different projection is possible, but that is also true.

216

CHAPTER 9. PROJECTIVE GEOMETRY

9.9.3 Combining 1-Dimensional Projections

What if we do one projection and follow
it by another, as in figure 9.21? Here we
have two projections. First, using P as the
center of projection, the points A, B, . . . ,
E are projected to A0 , B 0 , . . . , E 0 . Then
those primed points are projected to a
new line from a new center of projection,
Q to make A00 , B 00 , . . . , E 00 . This operation could be repeated as many times as
desired.

A
B
C
D
E

A
B
C

D
E

B C
D E
It should be clear that the operation
of repeated projection is equivalent to the
composition of the mathematical repreFigure 9.21: Combining 1-Dimensional
sentations of the projections. For examProjections
ple, if the projection centered at point P
Projective/Concatenate.T [M]
from line l1 to line l2 is represented by
f (x) and the projection centered at point
Q from line l2 to line l3 is represented by g(x), then the combinationproject from P ,
then from Qwill be represented by the composition of functions: g(f (x))first we
apply f , then we apply g to the result of applying f .
To take a concrete example, if the first projection is f (x) = x + 5 and the second
projection is g(x) = 1/x (both of which we know to be possible projections), the
resulting projection is
g(f (x)) = g(x + 5) = 1/(x + 5).
We already know how to make projections for f (x) = x, g(x) = x + , and
h(x) = 1/x, for any real 6= 0, and for any real . If we combine these in various
ways, how complicated an expression can we get? The answer is that starting from
only those three functions, we can make any function of the form:
f (x) =

x +
,
x +

where 6= 0, and we cannot get anything more complicated.

We will leave it as an (easy) exercise to show that the function above can be generated from the three primitive operations, but we will show that nothing more complicated can be generated. To see that we cannot get more complicated, let
f (x) =

x +
x +

(9.7)

g(x) =

x +
,
x +

(9.8)

and

9.9. HIGHER DIMENSIONAL PROJECTIVE GEOMETRY

217

where 6= 0 and 6= 0.
Let us calculate (g(f (x)):



x+
x+ +


x+
x+ +

g(f (x)) =

(x + ) + (x + )
(x + ) + (x + )
( + )x + ( + )
,
( + )x + ( + )

=
=

Which is of the same form, since all the sums of products in parentheses are just real
numbers.
The only thing that should be shown is the condition about a non-zero condition
like the original 6= 0.
For the final expression, this amounts to showing that

( + )( + ) ( + )( + ) 6= 0,
which can be simplified as follows:
+ + +

6= 0,

( )( )

6= 0.

6= 0,

This last is clearly true since neither of the terms above can be zero since the definitions
of f and g require that each of the terms be non-zero.
It turns out that every one-dimensional projection can be expressed in this form. In
fact, the form is a little over-specified. If
f (x) =

x +
x +

then multiplying , , , and all by the same constant non-zero number yields exactly
the same projection. But since any one of the four numbers can be zero we cannot
arbitrarily say that , for example, is 1.

9.9.4  Matrix Form in One Dimension

If you know something about matrix algebra, the calculation above should be pretty
familiar. Assuming that f and g are defined as in equations 9.7 and 9.8, take the

218

CHAPTER 9. PROJECTIVE GEOMETRY

numbers in the definitions of f and g and put them in matrices as follows 6 and multiply
them together as follows:


 



+ +
=
.


+ +

The matrix on the right above corresponds exactly to the composition of g(f (x)).
But there is even more to the matrix notation. Following what we did in two dimensions in Section 9.8, we can define one-dimensional homogeneous coordinates as
follows. The Euclidean point x is represented by (x, ), where 6= 0, and the point
at infinity will be represented by (, 0), where 6= 0. In other words, if 6= 0, then the
point with homogeneous coordinates (, ) corresponds to the Euclidean point /.
Now if we write the homogeneous coordinates of a point as a row vector, we can
represent the operation of projection as a matrix multiplication. Let f be defined as in
equation 9.7, and let it operate on a point with Euclidean coordinate x (which we will
represent as the row vector (x 1). The matrix multiplication gives:






x 1
= x + x + ,

but the point having the homogeneous coordinates in the vector on the right corresponds to the Euclidean point
x +
,
x +
as long as the denominator is non-zero, which is exactly what we wanted.
What all of the above shows is that by simply putting the coefficients of the projections into matrices, we can represent the operation of repeated projections as a product
of matrices. Similarly, the actual projection can be interpreted as the multiplication of
a vector by the projection matrix.
Just as both the homogeneous coordinates of a point may be multiplied by the
same non-zero constant and the point will remain the same, all the elements in any
projection matrix can be multiplied by the same non-zero constant and it will continue
to represent the same projection. This will also hold in any number of dimensionssee
Section 9.9.5.

9.9.5  Matrix Form in n Dimensions

The other nice thing about the matrix representation is that it extends trivially to higher
dimensional projective spaces. In two dimensions, for example, any projection or combination of projections can be represented as a 3 3 matrix such as:

 ,
(9.9)

6 Notice that the rows and columns are swappedin other words, although the numerator (the first row)
of the expression for f contains x + , the and are written in the first column of the matrix. They
could have been written as rows, but then the notation would be incompatible with the way that Geometer
represents transformations.

9.9. HIGHER DIMENSIONAL PROJECTIVE GEOMETRY

219

where the matrix above is non-singular. (The condition 6= 0 was sufficient

to assure a non-singular matrix in the 2 2 case, but it is more complicated in higher
dimensions7.) The composition of two such projections corresponds to the product of
their matrix representations.
If (xw, yw, w) is the homogeneous representation of a two-dimensional projective
point, then the operation of projection by the projection represented in 9.9 is achieved
by multiplication:

xw yw w  .

The pattern continuesin three dimensions, we need a 4 4 matrix, and in n
dimensions, an (n + 1) (n + 1) matrix.

This also shows why the images of three points are required to completely determine a one-dimensional projection, and why four points are required in two dimensions, five in three, and so on. The (n + 1) (n + 1) matrix has (n + 1) 2 = n2 + 2n + 1
entries, but since any constant multiple of the matrix is equivalent, there are basically
only (n + 1)2 1 = n2 + 2n = n(n + 2) values that can be chosen freely. In ndimensional projective space, each image of a point adds n constraints that must be
satisfied, or equivalently, n additional equations that must be satisfied. Since there are
n(n + 2) variables, the images of n + 2 points are required to completely determine the
projection. If n = 1 (one dimension), then n + 2 = 3 points are required, and so on.
One other nice feature of the matrix representation is that if you want to invert a
projection, you simply need to invert the matrix. In other words, if you have the matrix
to project from line 1 to line 2, and you want the matrix that undoes that operation
and does the reverse projection from line 2 to line 1, simply take the inverse of the
matrix. It is clear this will work, since a matrix times its inverse is the identity matrix
and multiplying the identity matrix by anything has no effect. That means if you apply
the first projection followed by the second, you will come exactly back to where you
started which is exactly what is meant by undoing the projection.

9.9.6 Points and Lines in Geometer

Recall in Section 9.8 that the coordinates for a general (two-dimensional) projective
point were (a, b, c) and for a line, we wrote them as [A, B, C] where at least one of the
numbers in both cases was non-zero. In fact, instead of being written [A, B, C], we
should more properly have written a line as a column vector since we have designated
points to be row vectors. It takes too much space to write column vectors as a stack
of three numbers in text, so it is often written as (A B C)T , where the T exponent
stands for transpose.
With lines written as column vectors, the transformation of lines by a matrix is
again a matrix multiplication. If (A B C)T represents a line and (A0 B 0 C 0 )T is the
7 Well,

not that complicatedthe condition is simply that the determinant be non-zero.

220

CHAPTER 9. PROJECTIVE GEOMETRY

result of transforming it by the matrix above, then:

0

A
A
 B = B0 .

C
C0

The point (x y 1) lies on the line (A B C) if and only if the product of the row
matrix and column matrix is zero:

A


x y 1 B = (Ax + By + C) = (0).
C

If calculations are done this way in homogeneous coordinates, there is no need ever
to check to see if lines are parallel or if points lie at infinity. The right thing happens in
every case. If Geometer needs to find the intersection of two lines, it blindly performs
the cross product operation described in Section 9.8 which works whether the lines be
parallel or not. If they were parallel, the result is simply the correct point at infinity
that can be used in further calculations.

9.9.7 Transformations Used by Geometer

In Geometers transformation commands, and even when operations like the rotation
of all the primitives are performed, all the transformations are achieved by multiplying
the coordinates by a suitable 3 3 matrix after having converted them to homogeneous
coordinates with the addition of a 1 as the third coordinate. Here are the basic matrices used by Geometer to transform points. In the equations below, P = (x y 1) is
the input vector:

1 0 0
(9.10)
P 0 1 0 = (x + tx y + ty 1)
tx ty 1

sx 0 0
P 0 sy 0 = (sx x sy y 1)
(9.11)
0 0 1

cos
sin 0
(9.12)
P sin cos 0 = (x cos y sin x sin + y cos 1)
0
0
1

Transformation 9.10 translates a point by tx in the x-direction and by ty in the y

direction. Transformation 9.11 scales a point by a factor of sx in the x-direction and
by sy in the y-direction. Finally, transformation 9.12 rotates a point counter-clockwise
by an angle of about the origin.
Combinations of these can be combined simply by multiplying the appropriate matrices together. For example, to rotate about the point (x0 , y0 ) one could simply translate by (x0 , y0 ), bringing the point (x0 , y0 ) to the origin. Then a rotation is applied,
and finally a translation by (+x0 , +y0 ) to return the origin back to its original location.

9.9. HIGHER DIMENSIONAL PROJECTIVE GEOMETRY

221

9.9.8  Constructing a Projection Matrix

In Section 9.9.2 we stated that if you know the images of any three distinct points in
the one-dimensional case, the projection is completely determined. So (using homogeneous coordinates), suppose we would like to find the matrix form of a projection f
that does the following:
f (x1 , w1 ) = (X1 , W1 )
f (x2 , w2 ) = (X2 , W2 )
f (x3 , w3 ) = (X3 , W3 )
We want to find a matrix

satisfying these three equations:

(x1

w1 )

(x2

w2 )

(x3

w3 )

= (X1

W1 ),

= (X2

W2 ),

= (X3

W3 ).

Well, not quite. If we could solve the equations above, that would be great, but
sometimes we cannot. The reason is, to take the first equation as an example, that
(X1 , W1 ) is equivalent to (k1 X1 , k1 W1 ) for any non-zero k1 . Variables k2 and k3
could similarly be inserted in the other equations.
Adding the ki terms, and re-writing the three equations above as a single matrix
equation, this is basically the system we are trying to solve:

x1
x2
x3


w1

w2

w3

k 1 X1
= k 2 X2
k 3 X3

k 1 W1
k 2 W2 ,
k 3 W3

where the twelve known values are the xi , wi , Xi , and Wi , and the unknowns are , ,
, , and the three ki .
If you multiply the whole mess out, you get six equations and 7 unknowns which
is usually bad news, but in our case, we can lock down one of the k i to be 1 since that
would be equivalent to multiplying the whole system by a constant, and then we can
solve the six equations and six unknowns in the usual (ugly) way.
A nicer way to do it is the following. It is not too hard to come up with the matrix
if the input values are simple. A great set of simple input values are these: (x 1 , w1 ) =
(1, 0), (x2 , w2 ) = (0, 1), and (x3 , w3 ) = (1, 1). All the ones and zeros make the
calculation quite a bit easier, and it is not hard to get the desired matrix.

CHAPTER 9. PROJECTIVE GEOMETRY

222

If you can do this, you can find the matrix form for a projection that takes the
points 0, 1, and to three arbitrary points. But the original problem required a matrix
representing the projection from any three points to any other three points. Well, we
can find the two transformations that take 0, 1, and to either the original or final set
of points, invert one of them, and multiply the results get the matrix we want.
In other words, suppose we want to map P1 to Q1 , P2 to Q2 , and P3 to Q3 where
the Pi and Qi are any points represented in homogeneous coordinates. Suppose we
find a matrix M such that M takes 0 to P1 , 1 to P2 , and to P3 . Suppose further that
N is a different matrix that takes 0 to Q1 , 1 to Q2 , and to Q3 . M 1 (M inverse)
takes P1 to 0, P2 to 1, and P3 to (it undoes the operation of M ).

If we apply M 1 first, then N , we take P1 first to 0, then to Q1 . Similarly, P2 goes

to 1, then to Q2 , and P3 goes to , and then to Q3 . The matrices M 1 and N are
combined with a simple matrix multiplication.
This may seem ugly, and it is probably overkill for the one-dimensional case, but
in 2, 3, or more dimensions, it is by far the best approach.
In 3 dimensions, for example, a transformation is completely determined by the
images of 5 points, so there are basically 20 equations and 20 unknowns (16 unknowns
in the 4-by-4 matrix plus the 5 ki values). But the problem becomes tractable if we find
the matrices for a particularly simple set of points: (1, 0, 0, 0), (0, 1, 0, 0), (0, 0, 1, 0),
(0, 0, 0, 1), and (1, 1, 1, 1). Find two such transformations and combine one with the
inverse of the other to get the desired general transformation.
If you just look at the matrices, it is easy to see why this approach works well. We
will show it here for two-dimensional projective space which should make the general
approach obvious.
Given 4 points (x1 , y1 , w1 ), . . . , (x4 , y4 , w4 ), we need to find the 3-by-3 matrix
[aij ] and constants k1 , k2 , k3 , and k4 such that:

1
0

0
1

0
1
0
1

0
a11
0
a21
1
a31
1

a12
a22
a32

a13
a23
a33

k 1 x1
k 2 x2
=
k 3 x3
k 4 x4

k1 y1
k2 y2
k3 y3
k4 y4

k 1 w1
k 2 w2
.
k 3 w3
k 4 w4

If we just multiply this out, we obtain the following 12 equations:

a11 = k1 x1
a21 = k2 x2
a31 = k3 x3
a11 + a21 + a31 = k4 x4

a12 = k1 y1
a22 = k2 y2
a32 = k3 y3
a12 + a22 + a32 = k4 y4

a13 = k1 w1
a23 = k2 w2
a33 = k3 w3
a13 + a23 + a33 = k4 w4

The first three rows of equations tell us that the aij are just multiples of the xi , yi ,
and wi , so the only real unknowns are the ki . We can let k4 = 1, so all we need to do
is solve this system:

9.10. THE EQUATION OF A CONIC

223

k 1 x1 + k 2 x2 + k 3 x3
k1 y1 + k 2 y2 + k 3 y3

= x4
= y4

k 1 w1 + k 2 w2 + k 3 w3

= w4 ,

which is just a straightforward system of three equations and three unknowns.

9.10 The Equation of a Conic

Pascals theorem provides a computationally efficient method to compute the classical
equation of a conic given any five distinct points that determine a conic. In classical
Euclidean geometry, the equation of a conic is given by:
Ax2 + Bxy + Cy 2 + Dx + Ey + F = 0.

(9.13)

Obviously, any multiple of this equation yields the same conic, but we cannot, for
example, divide through by A, since A might be zero. We are similarly prevented from
dividing through by any of the other coefficients.
Thus although there appear to be six degrees of freedom, there are only five. If we
know 5 distinct points on it, we should be able to determine all the points on a conic.
(We have to be careful of degeneraciesif any three of the points lie on a line, we will
get a degenerate conic.)
Suppose you are given the coordinates of five points on a conic: (x 0 , y0 ), (x1 , y1 ),
. . . , (x4 , y4 ). How can you determine A, B, C, D, E, and F in equation 9.13, above?
One way is that since you know the points lie on a conic, simply plug the values
(xi , yi ) into equation 9.13 and get five equations that look like:
Ax2i + Bxi yi + Cyi2 + Dxi + Eyi + F = 0.
Assume that F 6= 0 and divide through by F so the constant term is 1. Then
there are five equations and five unknowns, and these can be solved using standard
techniques from linear algebra. (Remember that the only variables are A, B, et cetera.
The terms with xi and yi are constants which you know.
If this succeeds, great. But what if F = 0? What if three of the points lie in a
straight line? Well, on a failure, we can then assume F = 0 and divide through by E
and see if we have better luck. If that fails, we can assume E = F = 0 and divide
through by D, et cetera. If all of those fail, then we know the set of points must contain
three in a line.
But if you are doing this on a computer, it is extremely unlikely that your solution
will failbecause of round-off errors and the finite precision of the computer, you will
get an answer as if F were something like 1015 a completely bogus result.
Pascals theorem provides a much better way.

224

CHAPTER 9. PROJECTIVE GEOMETRY

Given five points (xi , yi ), convert them to homogeneous coordinates: (xi , yi , 1).
Add a sixth unknown point that lies on the conic with coordinates (x, y, 1). Now write
down the equations that the six points must satisfy according to Pascals theorem. This
is easy to dowe know exactly how to find the equations of the lines that pass through
pairs of points, the equations of the points that lie on the intersections of those lines,
et cetera. For all the points but one, we have actual numbers. After you grind out the
conditions that must be satisfied by the coordinates of your sixth point, it will be in the
form of equation 9.13, and you can just read off A through F . If the solution fails for
some reason, it is not because of an equation with a zero coefficient; it is because three
or more of your given points lie on a straight line.

9.11 Finite Projective Planes

Although the typical computer geometry program will not help much with this subject,
it is interesting to note that the axioms of projective geometry nowhere say that there
must be an infinite number of points (or an infinite number of lines). There exist, in
fact, some finite geometries, one of which has as few as seven points and seven lines.
The smallest finite projective plane contains seven points and is called the Fano
Plane..
The seven points are A, B, C, D, E, F , and G, and the seven lines each contain
three points. Here are the lines: ADB, BEC, CF A, AGE, BGF , CGD, and DEF .
You can easily check that every two points determine a line, and every two lines intersect in exactly one point.
Clearly this is nothing like Euclidean geometryyou cannot draw it with straight
lines on a Euclidean plane.
For certain n 2 there exists a finite projective plane of with n2 + n + 1 points and
the same number of lines. Each line contains n + 1 points, and there are n + 1 lines
passing through each point. You may find it interesting to try to construct examples for
n = 3 (containing 13 points and lines) or even for n = 4 (with 21 points and lines).
There are still open questions about exactly which values of n 2 admit the existence of a finite projective plane. It is known that if n is a power of a prime number,
then a finite projective plane with n2 + n + 1 points and lines exists, but for other values of n nobody knows. No finite projective planes for such n have ever been found,
and for certain values of n the existence of such a projective plane has been proved
impossible.
Most small n are powers of primes: 2 = 21 , 3 = 31 , 4 = 22 , 5 = 51 , 7 = 71 ,
8 = 23 , 9 = 32 , and 11 = 111 . So among the first ten possibilities, only 6 and 10
are not powers of prime numbers. It has been proved that a projective plane of order
62 + 6 + 1 = 43 is impossible, but nobody knows whether or not a projective plane
corresponding to n = 10 (having 102 + 10 + 1 = 111 points and lines with 11 points
on each line and 11 lines going through each point) can exist. For more information,
see [Ryser, 1963].

9.11. FINITE PROJECTIVE PLANES

225

9.11.1 The Finite Projective Plane of Order 13

Here is a list of lines for such a plane, where the points are simply A, B, C, . . . , M ,
and the lines are denoted by l1 , l2 , l3 , . . . , l13 :
l1
l4
l7
l10

=
=
=
=

ABCD
ELJC
BF IL
AKLM

l2
l5
l8
l11
l13

=
=
=
=
=

EF GA
EIM D
CF HM
BGJM
DGHL

l3
l6
l9
l12

=
=
=
=

EHKB
AHIJ
DF JK
CGIK

You can check a few examples to see that every two points have one line in common
and every two lines have one point in common.

226

CHAPTER 9. PROJECTIVE GEOMETRY

Chapter

10

 Harmonic Point Sets

Harmonic point sets arise in a surprisingly large number of situations. The condition
of harmonicity is fundamentally a concept from projective geometry (see Chapter 9).
However, it is often easy to use in Euclidean situations, and it is also related to inversion
in a circle (see Chapter 8) and even to musical harmony!

10.1 What is an Harmonic Set?

Let us begin with a simple example. Consider the diagram in figure 10.1. Given
collinear points A, B and C, construct a circle with AB as its diameter. Choose a
point X arbitrarily on the circle and copy the angle AXC to the other side of the ray
XA. (This is most easily done in Geometer by reflecting the point C across the line
XA.) Let the line XC thus reflected across XA intersect the line AB at a point D.
This works whether C is between A and B or not, and only fails if C happens to be the
midpoint of the segment AB.
227

228

CHAPTER 10. HARMONIC POINT SETS

Notice that if you move the

point X around on the circle, the
location of D does not changeit
seems to depend only on the locations of points A, B and C. In
D
fact, if we move the pont X such
that the line DX is tangent to the
circle, it certainly appears that the
line XC is perpendicular to AB,
so it looks like points D and C are
inverses of each other relative to the circle.

Figure 10.1: Harmonic Division

Harmonic/BisectHarm.T [M]

10.1.1 Geometer Analysis of the Situation

If we test the diagram with Geometer, moving the points X and C during the test,
we obtain a list of relationships that are preserved, and in addition to the obvious ones
there are a few interesting ones as well:
Points [A B, C D] form an harmonic set.
Points C and D are inverse relative to circle c1.
Triangle area ratios:
[A C X] x [B D X] = [A D X] x [B C X]
Ratios:
[B C] x [X D] = [B D] x [X C]
[A C] x [X D] = [A D] x [X C]
[A C] x [B D] = [A D] x [B C]

The triangle area ratio is the same as the ratio of its bases where the bases are all
taken along the line DB with X at the vertex of all of them. Geometer also reports
that C and D are inverse relative to the circle which we suspected.
But the statement that the points [AB, CD] form an harmonic set is new. What
does it mean?

10.1.2 Definition of an Harmonic Set of Points

If the points A, B, C and D lie on a line, then the points are said to form an harmonic
set (written [AB, CD]) if:
AC BD = AD BC.

(10.1)

When this occurs, one of C or D will be interior to the segment AB and the other will
be outside as they are in figure 10.1.
If we rewrite equation 10.1 as ratios, it looks like this:
AD
AC
=
.
BC
BD

(10.2)

10.1. WHAT IS AN HARMONIC SET?

229

If we think of AB as a fixed segment, and if point C lies between A and B, then C

divides the segment into a certain ratio. What is interesting is that once C is chosen,
there exists another point D that also produces the same ratio, except that now one of
the lengths (AD in the case of figure 10.1) will be larger than the original segment.
Another way of saying this is that the points C and D divide the segment AB
internally and externally in the same ratio. Once the location of C or D is fixed, the
location of the other is determined. An interesting feature of this subdivision is that
if C and D divide the segment AB internally and externally in the same ratio, then
A and B divide the segment CD internally and externally in the same ratio, so the
relationship is symmetric. This is easy to see if we begin with the same equation 10.1
and divide by AD and BD instead to obtain:
BC
AC
=
,
AD
BD

(10.3)

which is exactly what is required to say that A and B divide the segment CD internally and externally in the same ratio. That is why the notation for an harmonic set is
symmetric. In fact, we have just proved that if [AB, CD] then [CD, AB].

10.1.3 Back to the Original Problem

The results we guessed after studying figure 10.1 are not hard to prove. An exercise in
Chapter 6 whose solution appears in the file Locus/Locus5.T proves that the locus of
points that lie in a fixed ratio of distances from two given points is a circle, so if the
circle passes through A and B, the distances from any point X on the circle to C and
D will be in the same ratio as are the distances from A and from B.
If we wish to show that C and D are inverses of each other relative to the circle,
that is equivalent to showing that when XC AB the line DX is tangent to the circle
at X. The easiest way to proceed is to construct the line OX, where O is the center of
the circle.
If it is still not clear how to proceed, construct the Geometer diagram similar to
the figure (containing the segment OX, where O is the center of the circle), but with a
right angle at C and test the diagram. The large set of equal and supplementary angles
obtained should provide more than enough hints to arrive at a proof.

10.1.4 Harmonic Point Sets in Projective Geometry

Of course you can see that if C is very near the midpoint of the segment AB, then in
order for [AB, CD] to hold, the point D must be very far away. In fact, once C is
the midpoint, there is no Euclidean point D satisfying [AB, CD], but if we choose to
include the point at infinity on the line AB, the segment is effectively divided internally
and externally in a ratio of 1 : 1.
General projections do not, of course, preserve lengths, but (as we shall see later)
they do preserve sets of harmonic points. In other words, if 4 points on a line are in

CHAPTER 10. HARMONIC POINT SETS

230

m
A

n
B

n
A

Figure 10.2: Constructing Harmonic Points

Harmonic/HarmConstruct.T [M]

an harmonic relationship and they are projected to another line, the projected points
will remain in the same harmonic relationship. We will even discover a numerical
measurement called the cross-ratio that measures for any 4 points how close those
points are to being in an harmonic relationship. Even if the points are not in an harmonic relationship, the numerical value of this cross-ratio will also be preserved by
projections.
Since harmonicity is a projective concept, then because of duality, there must be a
way to assign a meaning to an harmonic set of lines. In fact, four lines are said to be
in an harmonic relation if they all pass through a single point and any line not passing
through that point intersects the four lines in an harmonic set of points. The term pencil
of lines means a set of lines passing through a single point.
If you believe that harmonic sets of points are preserved under projection, it is clear
that if one line intersects the harmonic pencil of lines, then all other lines will do the
same: the intersections are projected from one line to the other from the common point
of the four harmonic lines.

10.2 Some Constructions of Harmonic Sets

The first example in Section 10.1 provides a method to construct an harmonic set of
four points given three of them. In this section we will display a few other approaches.
The first two are very straightforward.
Figure 10.2 displays two related constructions of points C and D given A, B, and
the ratio m/n into which the segment is to be divided internally and externally. In the
two parts of the diagram the segments labeled m and n are parallel, as is the line
passing though C 0 in the upper diagram. It is a fairly simple exercise in similar triangles

10.2. SOME CONSTRUCTIONS OF HARMONIC SETS

231

X
A

C
Q

Figure 10.3: Interior and Exterior Tangents

Harmonic/InExt.T [M]

to prove that in both cases the appropriate harmonic division has been constructed:
AD
m
AC
=
=
BC
BD
n

and

A0 C 0
A0 D 0
m
= 0 0 = .
0
0
BC
BD
n

Note: it may be interesting to examine the Geometer file for figure 10.2 in a text
editor to see how two constructions can be synchronized as they are in this diagram.
Figure 10.3 shows another situation in which harmonic points arise. If A and B
are the centers of two circles that do not intersect nor lie one inside the other, the
intersections of their common interior and exterior tangent lines (C and D in the figure)
form an harmonic set with A and B. This is quite easy to prove if we simply label the
points of tangency as in the figure. By similar triangles, we obtain: AC/R = BC/r
and AD/R = BD/r, where R and r are the radii of the larger and smaller circles,
respectively, in the diagram. From this it is trivial to show that [AB, CD].
Although the figures and
X
constructions shown above look
simple, an actual construction
G
with straightedge and compass
F
would require quite a few auxilY
iary lines to construct all the necessary parallel lines. In addition,
this method definitely requires
A
C
B
D=D
a compass since some lengths
Y
F
are copied. The next construcG
tion can be done entirely with a
straightedge.
Figure 10.4 shows what is
X
perhaps the fastest way to find
the point D given A, B and C
Figure 10.4: Harmonic Division
so that [AB, CD]. In the figure
Harmonic/FreeConstruct.T
[M]
the method is applied twice (the
second time with primed point
names) to illustrate that it does yield the same result no matter how you select the
free points.
Given three points A, B and C on a line, select an arbitrary point X not on the line
and connect it with straight lines to A, B and C. Now select another arbitrary point Y

CHAPTER 10. HARMONIC POINT SETS

232

1/5 1/4

1/3

1/2

Figure 10.5: Harmonic Construction

Harmonic/Music.T [M]

G
I

1/4

1/2 1/3

1/5

1/6

1/7
x

1/8 1/9
x

1/10 1/11 1/12

x

Figure 10.6: Musical Notes and Harmonic Numbers

on the line XC that is different from X and from C. AY and BY intersect BX and
AX at points F and G, respectively. The required point D lies at the intersection of
GF and AB.
We will delay the proof that this construction does, in fact, yield an harmonic set
of points until Section 10.4.

10.3 Harmonic Points and Music

Figure 10.5 is a construction of the points with x-coordinates 1/2, 1/3, 1/4, 1/5, . . .
starting from the interval [0, 1]. Draw a rectangle whose base is [01] and its diagonals meet over the point 1/2. Draw the line connecting the upper left corner with the
new point 1/2 and it intersects the long diagonal from 0 over the point labeled 1/3.
Continue as in the diagram to obtain 1/n for any integer n.
Labeling the points as in figure 10.5, we obtain all of the following harmonic relationsips: [0(1/2), (1/3)1], [0(1/3), (1/4)1], . . . [0(1/n), (1/(n + 1))1], with the ratios
into which the segments are being divided internally and externally being 1 : 2, 1 : 3,
1 : 4, et cetera. Simple arithmetic shows that this is true, but can you show that the
construction does in fact yield those ratios? Do not be surprised if Geometer is a little sluggish if you ask it to test the diagram Harmonic/Music.T; the list it generates
contains 63,129 lines of text specifying relationships that are satisfied!
You may have heard of the harmonic series which is simply the diverging sum:
1 + 1/2 + 1/3 + 1/4 + . . .. The relationships illustrated by the diagram show why the
name may be appropriate.

10.4. THE COMPLETE QUADRILATERAL

233

Figure 10.6 illustrates the relationship between the harmonic numbers and musical
notes. If a taut string, when plucked, sounds the low C note (labeled with a 1 above
the note in the figure), then if the same string has a node at its center, the note produced
will be the C one octave higher (labeled 1/2 in the figure). Similarly, if the node is
1/3, 1/4, et cetera of the distance from the end, the notes produced are those indicated
in the figure by 1/3, 1/4, and so on. The notes with an x above them are not exact
the frequency at which the string will vibrate is not exactly the defined frequency for
the note.

10.4 The Complete Quadrilateral

The word quadrilateral means four sides, and a complete quadrilateral begins
with four lines shown in figure 10.7 (not including the dashed lines AH, IB and IE).
The quadrilateral is made up of the solid lines AB, BC, CD and DA. The only
constraint on the lines is that no three of them pass through the same point. This
includes points at infinity, in the projective sensein other words, in addition to no
three of the lines passing though a point, no set of three of the lines may be parallel,
either1 .
These four lines will meet at six
points (A, B, C, D, E and F in the
A
figure). Three pairs of those points
are opposite each other: A is opB
posite C, B is opposite D and E
G
D C
is opposite F . The lines connecting these pairs of points (the dashed
I
F
H
E
lines in the diagram) are called di-

agonal lines. If we draw all three

diagonals, there will be three addiFigure 10.7: The Complete Quadrilateral
tional points where the diagonals inHarmonic/CompleteQuad.T [M]
tersect, and in the figure those are labeled G, H and I.
From the point of view of harmonic point sets, complete quadrilaterals are very
interesting. If you test the diagram in figure 10.7, the following results are obtained:
Points [E F, H I] form an harmonic set.
Points [B D, G I] form an harmonic set.
Points [A C, G H] form an harmonic set.

Notice also that the right thing happens if a pair of the lines in the quadrangle are
parallel. Suppose point C is moved2 so that AB k CD. It is an interesting (but not
1 Since the concept of a complete quadrilateral is a projective as well as a Euclidean notion, there should
be a dual concept and there is. The complete quadrangle is the figure formed from four arbitrary points
and the six lines that join them.
2 To avoid many other extraneous relationships in the diagram, the points that control the locations of the
lines in figure 10.7 are invisible, so to move them you will need to click on the button Show Invis to see and
manipulate them.

CHAPTER 10. HARMONIC POINT SETS

234

too difficult) exercise in Euclidean geometry to show that when AB k CD, the line
IEF H is parallel to both of them and IE = EH, so [IH, E] (where is the
point at infinity in the direction of that line). In other words, to maintain harmonicity,
the point F must move to infinity in the direction of the parallel lines AB and CD.
There is a great deal more harmonicity hidden in figure 10.7. If we simply connect
I with C (intersecting AF and AE at N and M ) and connect E with G (intersecting
F B at K and F A at L), then testing the resulting diagram yields:
Points
Points
Points
Points
Points
Points
Points
Points
Points

[E
[E
[C
[B
[B
[A
[A
[A
[A

G,
F,
I,
D,
C,
N,
D,
C,
M,

K
H
M
G
F
F
F
G
E

L]
I]
N]
I]
K]
D]
L]
H]
B]

form
form
form
form
form
form
form
form
form

an
an
an
an
an
an
an
an
an

harmonic
harmonic
harmonic
harmonic
harmonic
harmonic
harmonic
harmonic
harmonic

set.
set.
set.
set.
set.
set.
set.
set.
set.

If we are going to prove, for example, that [EF, HI] in figure 10.7 we are going to
need a bit more than what Geometers Test Diagram tells us. One nice way to proceed
is to look at the relationships that hold for a single point that divides a segment.
In figure 10.8 we have a segment
AB that is divided (in this case interP
nally) by a point C. In the complete
quadrilateral, all three points are con
nected to another one (called P in this
a
b x
example).
We would like to find an alternative
A
C
B
expression for AC/BC involving the
other sides of the triangle or the angles
Figure 10.8: Dividing a Segment
or something. In the figure, the other
Harmonic/SegDivide.D [D]
three sides have lengths a, b and x, and
the angles are named and . If we
apply the law of sines to the triangles 4AP C and 4BP C we obtain:
x
AC
=
sin
sin A

and

BC
x
=
.
sin
sin B

(10.4)

If we combine the two equations in 10.4 in such a way that we can eliminate the variable
x, we obtain:
AC
sin sin B
=
(10.5)
BC
sin sin A
Note that this relationship holds whether C is internal or external to the segment AB.
If there is another point D dividing AB, we can obtain a similar expression by
replacing C by D, by and by in equation 10.5, where and are the angles
made by the line P D with the sides P B and P A, respectively. Here is the result:
sin sin B
AD
=
BD
sin sin A

(10.6)

10.4. THE COMPLETE QUADRILATERAL

235

If [AB, CD] then (AC/BC = AD/BD, or (AC BD)/(AD BC) = 1, so let

us calculate the ratio (AC BD)/(AD BC) from equations 10.5 and 10.6:
sin sin
AC BD
=
.
AD BC
sin sin

(10.7)

Notice that the expression on the right side of equation 10.7 depends only on the
angles made by the sides P C and P D relative to the sides P A and P B. This means
that once the four lines that meet at P are determined, any line not passing through P
will have the same value of the expression on the right of equation 10.7 as any other. In
other words, the value of that expression is preserved by projection onto another line.
If it is 1, meaning that the four points are in a harmonic configuration, they will remain
harmonic after a projection, and if they were not, they can never be projected into an
harmonic configuration.
The quantity (AC BD)/(AD BC) is very important since it is not only preserved by projection, but its value is 1 exactly when the four points are in an harmonic
configuration. It is called the cross ratio of the four points3 . The cross ratio is usually
written: {AB, CD} = (AC BD)/(AD BC).
If we return to figure 10.1 which is the first time we encountered points in an
harmonic ratio, it is easy to prove that the construction there generates an harmonic
set. In that figure, let = DXA = CXA and let = BXC. We know that
+ = 90 . If we substitute these angles into the expression on the right of equation 10.7 we obtain:
sin sin(90 + )
sin sin
sin sin
AC BD
=
=
=
= 1.
AD BC
sin sin()
sin(90 )( sin )
( sin )( sin )
Since the cross ratio is preserved under projection, we only need to prove that the
various harmonic ratios observed in the complete quadrilateral hold in a single configuration. This is because any configuration can be projected to any other (four points can
be mapped to four other arbitrary points, assuming the the initial and final configurations are non-degeneratesee Section 9.9.8). We can choose that single configuration
any way we want. Clearly we could choose a nice Euclidean configuration and calculate all the distances, so the result can be proved completely in Euclidean geometry, but
if we are willing to look at the result in a projective space, the result becomes almost
trivial.
In figure 10.7, for example, if ABCD is a square, then clearly G will be at the
center, so AG will meet EF at a point H that is the midpoint of F E. The segment
DB will be parallel to F E, so the point I will become the point at infinity. Thus the
points I, F , H and E satisfy [EF, HI] and so they will for any configuration of A, B,
C and D. The harmonicity of [BD, GI] and [AC, GH] can be proved similarly.
A quick comparison of figures 10.7 and 10.4 should make it obvious why the
straightedge-only construction worksit just constructs a complete quadrilateral with
A, B and C suitably placed.
3 Sometimes directed segments are used in the definition of the cross ratio, and if so, the statement that
the four points lie in an harmonic configuration replaces the value 1 by 1

236

CHAPTER 10. HARMONIC POINT SETS

One final note about the cross ratio is this: If we know the cross ratio of four points
in a particular order, the cross ratios of all other possible arrangements of those points
are determined. A little algebra shows, for example, that if {AB, CD} = k then
{BA, CD} = 1/k, {AC, BD} = 1 + k, and so on.

10.5 Poles and Polars

We noted earlier that there seems to be a relationship between inversion in a circle and
an harmonic set of points. In fact, in figure 10.1 the points D and C are inverses relative
to the circle with diameter AB. Inversion is very powerful in Euclidean geometry so
it is reasonable to search for a similar operation in projective geometry where this new
operation will, of course, have a dual operation where the roles of points and lines
are interchanged.
This is accomplished with poles and polars but instead of mapping a point to its
inverse point, a point is mapped to its polar which is a line that is reciprocal to that
point. The reciprocal of a line is its pole which is the point whose polar is the line.
In other words, the projective version of inversion will associate each point with a line
and vice-versa in a manner that is similar to inversion in Euclidean geometry.
Of course in projective geometry
there is no such thing as a circle, and
a full-blown projective definition of
poles and polars must apply to a general conic section, but let us begin by
looking at poles and polars in the Euclidean plane where this new sort of inversion will be relative to a circle.

Figure 10.9 shows the construcFigure 10.9: Pole and Polar

tion. Given a circle and a point A,
Harmonic/PolePolar.D [M]
let A0 be the inverse (point) of A relative to that circle. Now construct a line
through A0 perpendicular to the line AA0 . This line is the polar of A, and A is the pole
of this line. Thus there is a polar for every point except the center of the circle (and in
projective geometry, of course, the polar of the center is the line at infinity). Similarly,
every line has a pole except for the lines that pass through the center of the circle (and
in projective geometry those lines have poles at infinity in the direction perpendicular
to each line).
At first it seems that there is very little difference between the concept of pole and
polar versus the Euclidean inverse in a circle. Figure 10.10 illustrates the result of
finding the polars of a set of evenly-spaced points around a circle relative to a fixed
circle in the center of each of the four examples. From the figure it is obvious that the
polars of the points in a circle do not form another circle (or straight line) as they do
when those same points are subjected to the standard Euclidean inversion. Figure 10.10
illustrates only a small number of examples. The Geometer diagram allows you to

10.5. POLES AND POLARS

237

r
A
A

Figure 10.10: Polars of Circles

Harmonic/PolarInverse.T [S]

A=A

A
B

B
Figure 10.11: Poles, Polars, and Conjugates
Harmonic/PolePolar1.D [M]

adjust the center of the circle of points whose polars are to be drawn and the length of
the line segment on the left controls the radius of that circle. When the center and radius
are set, the Run Script command draws a set of polars relative to 90 evenly-spaced
points around the circle. The Geometer script itself perhaps merits an examination in
a text editor to see exactly how it works.
Here are a few basic properties of Euclidean (as opposed to projective) poles and
polars as defined relative to a circle (instead of relative to a general conic section). None
of them is particularly hard to prove, and most depend on the properties of inversion in
a circle.
1. If A lies on the circle, A = A0 and the polar of A will pass through A and be
tangent to the circle at that point. See the left diagram in figure 10.11.
2. If B is a point on the polar of point A, then A lies on the polar of B. In this
situation, A and B are called conjugate points and their polars are conjugate
lines. See the right diagram in figure 10.11.
3. If a secant through a circle is drawn through a point P and tangents are drawn
at the points where the secant intersects the circle, then the intersection of those
tangent lines lies on the polar of P . If this process is repeated, the two points
found on the polar can be used to construct the polar. See figure 10.12 where the
intersections of the tangents at Q0 and Q00 determine the polar of the point P .
4. If the line joining two conjugate points intersects the circle, the two intersections
of that line with the circle and the two conjugate points form an harmonic set.
5. See figure 10.13. If two secants are drawn from a point P that is exterior to the
circle and the complete quadrangle is formed from the four intersections (A 0 , A00 ,

CHAPTER 10. HARMONIC POINT SETS

238

P
A A

Q
B

B
P

Figure 10.12: Constructing a Polar

Harmonic/PolePolar2.D [M]

P
A A
T Q T

B
P

B
P

Figure 10.13: Polar and Tangent Construction

Harmonic/PolePolar3.D [M]

10.5. POLES AND POLARS

239

A
C
A C
B

Figure 10.14: The Polar Circle

Harmonic/PolarCircle.T [M]

B 0 and B 00 ) of the secants with the circle, then the two additional intersections
in the complete quadrangle (Q0 and Q00 ) lie on the polar of the point P . This is a
straightedge-only construction of the polar. Furthermore, the intersections of the
polar with the circle when connected to P (T 0 and T 00 ) are the two tangents to
the circle passing through P , so tangents to a circle can also be constructed with
a straightedge only.
6. The angle between the polars of A and B is the same as the angle AOB, where
O is the center of the circle.
7. The distances from the center of a circle to any two points is proportional to the
distances of each point from the polar of the other.
8. If two circles are orthogonal, then the (extension of a) diameter of one will cut
the other in two points that are harmonic relative to the endpoints of the diameter
in the first, and conversely.
9. A triangle is called self conjugate relative to a circle if its sides are the polars
of its vertices. A self conjugate triangle can be constructed by taking an arbitrary
point other than the center of the circle as one vertex, then choosing another
point on its polar as the second vertex, and the third vertex at the intersection
of the polars of the first two. Such a triangle must contain an obtuse angle. In
figure 10.14, the triangle 4ABC is self conjugate since the polars of points A,
B and C are the lines BC, CA and AB, respectively.
10. If a triangle is self conjugate relative to a circle, that circle is called the polar
circle of the triangle. Any obtuse triangle has a polar circle. In figure 10.14 the
circle is the polar circle of 4ABC.
11. The nine-point circle and the circumcircle are inverses of each other relative to
the polar circle. This is obvious since the circumcircle passes through the
vertices of a triangle 4ABC and the nine-point circle passes through the feet of
the altitudes, which are the inverses of the vertices A, B and C. Figure 10.14
makes it clear exactly what is going on.
12. If a complete quadrangle ABCD is inscribed in a circle, the intersections of
the lines AB with CD, AC with BD and AD with BC form a self-conjugate
triangle. See figure 10.15.

240

CHAPTER 10. HARMONIC POINT SETS

Figure 10.15: Constructing a Self Conjugate Triangle

Harmonic/SelfConj.T [M]

Figure 10.16: Apollonius CCC problem

Harmonic/App10.T [M]

13. The altitudes of a self-conjugate triangle pass through the center of the circle.
14. The polar of a point A relative to a circle with center O is the radical axis of that
circle with the circle having diameter OA.
15. The reciprocal of any rectangle with the same center as the circle is a rhombus.

10.6 The Problem of Apollonius (Again)

In Section 8.15 we presented a solution, using inversion in a circle, to the problem of
constructing all the circles that are simultaneously tangent to three given circles which
we labeled there as CCC. The solution there required that certain decisions be made
during the construction. For example, the first step was to decide which of the three
circles had the smallest radius so that that length could be subtracted from the other
two to reduce the problem to one of finding the circles tangent to two given circles and
passing through a point.
The method works fine, but it is difficult to produce a Geometer diagram that works
for arbitrary circles using this method. In fact, in figure 8.25 the centers and radii of
the three given circles can only be moved a little bit before the Geometer diagram fails
to work.
Using poles of a circle, there is a mechanical construction of the tangent circles due
to Gergonne which we present here without proof.

10.7. POLES AND POLARS RELATIVE TO CONICS

241

For each pair of circles, find the two homothetic centers. The homothetic centers
are the points from which one circle is a projection of the other. If the circles are
disjoint, for example, the homothetic centers are the points where the internal tangents
intersect and where the external tangents intersect.
These six points lie on four lines. For each of those lines, find the poles of that line
relative to each of the three circles, and if they all exist, connect them to the radical
center of the three circles. The lines connecting the poles to the radical center will
intersect each circle twice, and those six points are the points of tangency of two of the
solutions to the CCC Apollonius problem. Since there are four lines, there are up to
eight solutions. The Geometer file Harmonic/App10.T contains that construction, and
is illustrated in figure 10.16.

10.7 Poles and Polars Relative to Conics

We cannot construct poles and polars for
a general conic using the Euclidean conA
cept of inversion, but a construction like
that used in example 5 in the previous
section will work for any conic section
not just a circle.
It is surprisingly difficult to construct
Figure 10.17: Projective Pole and Polar
a Geometer diagram that will produce
Harmonic/ProjPolar.T [M]
the polar for any pole and any conic section as illustrated by the example in figure 10.17. In that figure (and in the Geometer diagram) the points indicated with solid
diamonds control the shape of the conic section. They can be moved as can the point
A. The line is the polar of A.
With a fixed point and a fixed conic, it is easybegin by drawing two lines that pass
through the conic in two different places and then continue exactly as was done with
the circle. The problem, of course, is how does one draw two lines that are guaranteed
to be different and to intersect the conic without knowing ahead of time what the conic
looks like and where the point is relative to it? It is a very good exercise to try to make
a diagram like that in figure 10.17 before reading the following material.
We can easily obtain one point (say P ) on the conic using Geometers P on Conic
command. The line connecting A and P will intersect the conic in another point,
but even that is hard to pick consistently. If you simply choose the other point of
intersection Q of the line AP with the conic using Geometers LCon=>P command it
will work fine for the initial construction, but as you change the shape of the conic and
the location of A you may discover that suddenly P and Q are the same point so the
line passing through the pair of them will behave erratically, to say the least!
What happens is that upon selecting the line and conic when you use the LCon=>P
command, Geometer finds the point nearest the last location of the mouse when you
clicked on the conic. Geometer solves the line-conic intersection problem by solving a quadratic equation that has two roots, depending upon whether the sign of the

242

CHAPTER 10. HARMONIC POINT SETS

discriminant is positive or negative. If it is positive for for the point of intersection,

Geometer always uses the positive value of the discriminant for that solution and viceversa. The new location of the point P on the conic, however, is constantly reprojected
to the conic as A moves or as the shape of the conic changes. The location of P may
suddenly become the same as Q because of that.
The best solution is that once you have the line AP that intersects the conic twice to
change P to the invisible color, and then create two new points at the two intersections
of the line with the conic. Geometer will find that it uses a positive discriminant for
one and a negative discriminant for the other, so even if P swaps to a different solution,
the two intersections will remain different.
Thus it is easy to obtain one line and its two different intersections with the conic.
How do you obtain another line and pair of points?
If you simply choose another point P 0 constrained to lie on the conic, as you move
things around and P and P 0 are constantly reprojected, it is overwhelmingly likely that
they will eventually converge to the same point. Try making a conic section through 5
points, constrain two points to lie on that conic and then change the shape of the conic
by moving the control points. You will find that almost always the two points will
converge to the same location on the conic. They may not be identical due to computer
round-off error, but the distance between them will be so tiny that the line that joins
them is very unstable.
But a perpendicular through P is guaranteed to intersect the conic in a new point,
and that new point is generally not near either of the two intersections of the line AP .
Of course this point has the same problem with respect to its definition and if we are
not careful, it may collapse to be the same as P . If you try the same trick, however,
that may not work, since P and this new point may swap, and your line through P and
A will, essentially, not pass through P .
Here is the solution: Pick P on the conic and draw the line AP . Through P ,
construct a perpendicular to AP . Make P invisible and find the two intersections of
the perpendicular line with the conic. When these different points are connected to A,
they are guaranteed to be different lines4 .
Draw the two lines from the intersections of the line perpendicular to AP at P .
Next, color the intersection points invisible, and make two new points on each line that
intersect the conic in different places. These four points can then be used to construct
the polar in the same manner that it was constructed for a circle.

4 Well, almost guaranteed: with incredibly bad luck, P may lie at a point where AP is tangent to the
conic and hence only intersects it in one place, but this is very, very unlikely.

Chapter

11

Geometric Presentations
If you are trying to solve a problem for yourself then as long as you understand what
you are doing, the beauty and neatness of the solution is not too important. But as soon
as you try to present your results to someone else it is important that the presentation
be neat and clear. If you are a geometry teacher using Geometer in front of a class, the
presentation will be easier for your students to understand if you make some effort to
polish each diagram before you present it.
In this chapter we will investigate methods to make diagrams easier to understand.
Teachers of geometry need to do this every day, but many mathematicians need to do
the same thing from time to time. In addition to dynamic presentations, Geometer can
also be quite useful for helping generate illustrations for publication, either electronically or printed on paper. Every illustration in this book (and every one in the reference
manual except for the screen shots) was prepared using Geometer.
The chapter is much more about techniques for using Geometer (or other computer
geometry programs) than it is about specific geometric facts, although the examples
243

244

CHAPTER 11. GEOMETRIC PRESENTATIONS

have been chosen to include some interesting results. Since many of the features and
methods described here are specific to Geometer you may wish to have the reference
guide available as you read this. Other computer geometry programs have most of
these features, but probably not all, and Geometer is missing some features found in
those programs.
The reference guide that comes with the Geometer release contains a tutorial for
teachers that is a useful addendum to the material in this chapter.

11.1 General Considerations

There are some obvious things to keep in mind, no matter whether your presentation
is electronic, on an overhead projector, written in chalk on a blackboard or printed in a
book or handout. Here are a few:
Use standard notation if possible. If that is impossible, at least use consistent
notation. In other words, use I for the incenter of a triangle, O for the
circumcenter and H for the orthocenter. If your triangle is 4ABC, let a, b and
c represent the lengths of the sides opposite the vertices A, B and C, respectively.
If you need to name angles, use Greek letters, and use , and for the angles
at vertices A, B and C, respectively. Use s for the semiperimeter, et cetera.
Make certain it is easy to read labels, especially the important ones. The nice
thing about Geometer diagrams is that you can usually adjust the diagram in
such a way that the labels do not fall on top of each other and so that they are not
obscured by other lines or circles in the diagram.
Use appropriate fonts and point sizes for the illustration. In a classroom the text
has to be big enough to be read from the back of the class. In a book, smaller
fonts may help remove the clutter from a diagram.
Show a general configuration if possible. In other words, if the theorem you
wish to demonstrate applies to all triangles, do not display it with a triangle that
is close to being equilateral or isosceles.
Eliminate extraneous information. Hide your construction lines by making them
invisible. Eliminate names from points that are not important to the presentation.
Do not color things differently unless you have a good reason to do so.
Do use colors, line stippling, angle markings, point shapes, et cetera when there
is a good reason. For example, if you have a diagram where you construct a
similar object based on each side of a triangle and it might be confusing what
corresponds to which side, draw the three sides in different colors, and draw
the objects associated with each side in the same color as the side. Similarly,
if you are using inversion in a circle, you might draw objects in the original
configuration in one color and in the inverted configuration in another.
In locus-of-points illustrations, try to make the representative points regularly
spaced.

11.2. IMPROVING THE APPEARANCE OF A DIAGRAM

245

11.2 Improving the Appearance of a Diagram

Many of Geometers primitives can be drawn in different styles. All can be drawn in
different colors, but the other styles are primitive-dependent.

11.2.1 Colors
In addition to the standard colors that are available in the color selection box in Geometers control area there are 24 additional available colors that you can customize
and use from Geometers text editor. The additional colors have names .c8, .c9, . . . ,
.c31. (Colors .c0 through .c7 are the reserved colors: black, red, green, yellow, et
cetera.)
In a Geometer file, the definition of the colors .c8 through .c31 can be set to a
particular combination of red, green and blue components that vary from 0.0 to 1.0.
To set the color .c8 to be pink and then to use it as the color of point p1, insert a line
defining the color components for .c8 and edit the line for p1 so that it looks something
like this:
.c8 = (1.000000, 0.400000, 0.400000);
p1 = .free(0, 0, .c8, "A");

The definition of color .c8 above will remain in effect only as long as that particular
Geometer file is loaded.

11.2.2 Names
Every primitive can have a user-assigned name, but by default, names are only assigned
to points as they are created, and this feature can be turned on and off in the Edit Preferences dialog. The name can be as simple as a single character or it can be a whole
string. To create or change the name of a primitive, select it and then use the Edit Name
command.
The names of points, lines and angles are drawn on the screen slightly above and to
the right of a point, or at the center of a line segment, either slightly above or slightly
below the segment depending upon whether the segment slopes up or down. The name
of an angle is drawn inside the angle. It is a bit difficult to draw angle names in exactly
the right place, so for high-quality drawings you will sometimes need to resort to other
methods to place angle names properly.
Another primitive whose name is used by Geometer is the floating-point number
(the .flt) that you can only have entered with the text editor. If a .flt has a name, the
value of that number is drawn in the upper left of the display window, together with the
name. For example, if the .flt named pi has the value 3.14159, it will be presented
as the string: pi = 3.142 in the upper left of the screen. All values are presented only
to three decimal places, since Geometer cannot guarantee accuracy much better than
that. If there is more than one named .flt, all the values are listed in a column.

246

CHAPTER 11. GEOMETRIC PRESENTATIONS

If you wish to display the length of a line segment, the area of a polygon, or the
measure of an angle, you can select that item and use the Display Value command. In
that case the primitives length, area, or measure is drawn in the same way in the upper
left portion of the display window. If the line segment named b has length 1.234 it will
be presented as b = 1.234. In the current version of Geometer, the coordinate system
runs from 1.0 to 1.0 in the shorter dimension of the viewing window so all lengths
and areas will be presented based on these dimensions.
The Display Value command is easy to useyou simply need to select a primitive
and toggle it on or offbut it is somewhat restricted by Geometers coordinate system.
For high-quality presentations you may need to resort to working directly with the
.flts. See Section 11.5.
Finally, the text you use for a primitive name can have certain special characters,
superscripts, subscripts, and so on. See Section 11.2.4.

11.2.3 Primitive-Specific Styles

In addition to colors and names, points, lines, polygons and angles can be displayed
in various modes. Points can be drawn as dots, small circles, diamonds, or as nothing
at all. Lines can be drawn with various stippling options, hash marks (to indicate
congruence to other segments), or as a segment, ray or full line. Different sorts of
angle markings are available, different polygon fill types are available, and the line
width of any line, circle, conic section or Bezier curve can be set from 0.5 to 6 pixels in
width. On the computer screen a width of 0.5 pixels cannot be shown (so it is displayed
as one pixel wide), but when printed it will produce a lighter line.
One of the main uses of a point with nothing displayed is for use as a label in a highquality illustration. Although no mark is made for the point, its name is displayed, so
it is basically a piece of text that can be moved to any position on the screen. If you
would like to have it locked at a certain location, use a .pinned point instead of a .free
point.
All these display styles for primitives are available in the Style pulldown menu or in
the selection boxes in the control area. Rather than give a complete description of each
here, it is easier just to draw some primitives with geometer, select them, and change
the styles of each to look at the various options.
In classroom presentations, sometimes the thin width of the lines in Geometer
makes them difficult to see across a classroom, especially if the image is projected on
a television screen instead of a computer screen projector. You can make the lines in
your diagram wider, of course, but then they will be inappropriate for close-up viewing.
The command Wide Lines (available in the Styles pulldown) doubles the width of all
the lines. The keyboard shortcut is Ctrl-L or Ctrl-l (lower-case L) which toggles the
wide lines on and off.
The only reason to use the polygon primitive in Geometer is to obtain its area
automatically or to have an easy way to stipple or fill regions for display purposes.

11.3. BUILDING GEOMETER PROOFS AND CONSTRUCTIONS

247

11.2.4 Descriptive Text

The text that appears at the bottom of the display window is the same as the text that
names primitives with the one exception that a primitive name must consist of a single
line of textno newline characters.
In addition to the standard characters, Geometer is capable of drawing a number
of special characters, subscripts, superscripts, and even lines or arcs over text. The
special characters are all introduced by a \ character. If you would like to talk about
4ABC, enter the following text: \triangleABC. The leading \ tells Geometer to
try to find a command name after it (and it finds triangle in this case). If it does find
a match, the matching text is replaced by the corresponding symbol. See the reference
manual for a complete set of special characters supported in this way.
Two other commonly-used text commands produce superscripts and subscripts.
The idea is similar: To produce A1 , type: A\sub{1}; to produce B 3 , type:
B\sup{3}. There are a few other special text commands that are rarely used, but
are documented in the reference manual.
Finally, Geometer has a somewhat unusual method for indicating newlines within
a piece of text in the textual description. If you type in your text and edit it using the
dialog box available under the Descriptive Text command in the Primitives pulldown
this is all handled automatically. If you are typing in text by hand using the editor, here
is what it looks like.
Suppose you would like your textual description to read like this with two lines of
text:
Congruent Triangles:
4ABC, 4DEF , 4GHI
This is exactly what you would type into the Geometer text file to obtain that result:
.text("Congruent Triangles:
\triangleABC, \triangleDEF, \triangleGHI");

Note that the opening and closing quotation marks are on different lines which is
very non-standard in computer languages. It is easier to type, however.

11.3 Building Geometer Proofs and Constructions

It takes a fair amount of work to build a beautiful proof or construction that is ideally
suited for presentation in a classroom. Once it is done, however, it can be used over
and over again. In this section we will build a few examples in a step by step manner.
In this section we assume that you know how to prove a result. You are simply
trying to present that proof in the clearest possible way.
Perhaps the best way to use this section is to read the description of how a particular
demonstration is built and then try to build it yourself using Geometer. If you have
problems, read the section again, presumably with much more interest.

CHAPTER 11. GEOMETRIC PRESENTATIONS

248

11.3.1 The Incenter

As a first example, let us build a proof of the fact that the three angle bisectors of a
triangle meet at a single point that is the center of the inscribed circle (the point is also
called the incenter) of the triangle.
Although we may have to use the text editor to put in the final touches, we would
like to do as much as possible using Geometers standard graphical user interface.
Usually the best way to construct a proof is to draw the whole thing at once, including
all construction lines that you want to be visible during the proof. Then return to the
individual steps of the proof, making lines, points, angles, et cetera, invisible if they
are not needed in that step and possibly emphasizing them if they are important.
Figure 11.1 displays all the visible
C
lines and points that we draw before beginning to construct the step by step proof.
Refer to this diagram in the discussion beA
low.
B
In the case of the incenter, we begin
I
by drawing an arbitrary triangle 4ABC
B E
where A, B and C are free points. Next,
D
C
construct two angle bisectors at angles A
A
and B using the PPP=>P Bis command to
find a point that bisects an angle defined
by three points. Since we know that the
Figure 11.1: Three Incenters Proof
three bisectors meet at a point (remember,
Presenting/ThreeIns.T [P]
we are only trying to present a proof, not
to find it), there is no need to construct the third bisecting point.
Any two angle bisectors could have been used, but it is conventional in cases like
that to use the angles with the names closest the beginning of the alphabet. In this
simple case it is hard to make a mistake, but if you had somehow made a mistake and
drawn an incorrect point and later erased it, your triangle might be 4ABD. Since it
is again conventional to use a logical set of names, if this occurs, simply change the
name of the point mislabeled D to be C.
In complex diagrams, it may be easiest just to turn off the automatic point naming
of Geometer and to fill in the names that you need later or as you go. In this case,
for example, the points that bisect the angles will automatically be assigned the names
D and E and we will have to remove those names later or make the points invisible.
The command to toggle on and off the automatic assignment of names is Point Names
under the Edit pulldown menu.
Draw the angle bisectors from A to D and from B to E and find their intersection
which we label as I since we know it will be the incenter.
At this point, since we are surely not going to need to see points D and E in the
final proof or the short segments AD and BE, it might be a good idea to erase them
by changing them to have the invisible color1 .
1 It

is a minor point, but it is a good idea to make segments AD and BE invisible even though the lines

11.3. BUILDING GEOMETER PROOFS AND CONSTRUCTIONS

249

WE complete the drawing of the bisectors by drawing the segments AI, BI and
CI.
Since we are going to want to indicate to the viewers that the lines AI, BI and
CI are angle bisectors, we should also draw in the six angles CAI, IAB, ABI,
IBC, BCI and ICA using the PPP=>A command. To indicate that there are
three pairs of equal angles, use the different angle styles to make them different. The
best way to do this is probably to use the angle styles with the slashes, since otherwise
it will not be obvious whether you are talking about the single angle CAB or the two
angles CAI and IAB. See the figure.
To complete the drawing we will need to drop perpendicular line segments to the
sides of 4ABC. Use the command PL=>L Perp to draw the three perpendiculars and
then use the LL=>P three times to create points at the feet of these perpendiculars. We
need points here since we will need to name those points in the textual presentation of
the proof. It might be nice to give those points logical names, so we use the Edit Name
command in the Edit pulldown menu to change the names to A 0 , B 0 and C 0 as in the
figure. The final touch is to draw the incircle centered at I and passing through A 0 (or
B 0 or C 0 ), or just draw the circle passing through the three points A0 , B 0 and C 0 .
Now that everything is drawn, let us convert our diagram to a proof. At this point,
every primitive is either visible on all the layers or invisible on all the layers. The ones
that are invisible at this point were made that way because they will never be shown,
so we do not need to worry about them. We will, however, need to make the visible
primitives change color and/or become invisible as we move from layer to layer in the
presentation of the proof.
A Geometer proof works as follows. When you press the Start button, only items
on layer zero are visible. Each time you press the Next button the visible layer advances
by one so after one press you will view items on layer one, et cetera. If you want
geometer to begin with only layer zero visible, you can add the line .l0; to the file.
This is commonly done on prepared proofs or constructions.
By default, the layer information is hidden from users, so if it is not visible you
probably want to turn it on with the Layer Display command in the Edit pulldown
menu. If you are going to produce a lot of proof or geometric construction demonstrations you can use the Edit Preferences entry in the Edit pulldown to have the layers
displayed every time you start the Geometer program.
An easy way to build the proof at this point is to press the Start button (so that
only layer zero can be viewed and edited) and then to make the diagram on that layer
look exactly as you want it to look. When that step of the proof is complete, press the
Next button to advance to the next step of the proof, and continue that way to the end.
Geometer has 32 layers, which should be sufficient for almost any demonstration.
In this section we will describe in detail the production of the first two steps of the
proof, but to complete it, you will have to perform the same general process on each
AI and BI will be drawn on top of them. This is because due to the fact that the lines are drawn as discrete
pixels, the part of the line AI that is supposed to be identical to AD will almost certainly differ in a few
pixels, so if AD is not made invisible, the line from A to I will be marginally wider during the part from A
to D. Also, if the triangle is made small enough, points D or E could be pushed outside of it.

250

CHAPTER 11. GEOMETRIC PRESENTATIONS

step. To view the final result, load the file Presenting/ThreeIns.T and press the Next
button to step through the entire proof.
In a computer geometry program it is probably best in the first step of the proof
to display the final result and explain it so that users can manipulate the figure to see
exactly what is happening. A clear statement of the theorem should also be presented
in that first step (on layer zero).
Save your diagram so far (as a sort of backup in case you make a terrible error) and
then click on the Start button. Under Layer Control only box zero should be checked
after doing this.
While viewing layer zero, select the points A0 , B 0 and C 0 and set them to the
invisible color. Similarly, set the line segments connecting I to those three points also
to be invisible. All the rest of the diagram should be visible since it demonstrates the
theorem we are going to prove.
Next, enter the description that is to appear on layer zero. It should probably say
something like this:
In \triangleABC the three internal angle bisectors
meet at a point I which is the center of the inscribed
circle--the circle inside the triangle that is
simultaneously tangent to all three sides. Move
points A, B and C.

You can enter this description by selecting the Descriptive Text entry under the
Text entry in the Primitives pulldown menu. If you prefer, you can enter it using the
Edit Geometery command also in the Edit pulldown menu, but in this case you will
have to include with it the wrappings that make it a text file command:
.text("In \triangleABC the three internal angle bisectors
meet at a point I which is the center of the inscribed
circle--the circle inside the triangle that is
simultaneously tangent to all three sides. Move
points A, B and C.", .l0);

It is nice to include in the description an indication to the user of what parts of the
diagram can be manipulated. In this case, for example, the user cannot move point
Ionly points A, B and C are free.
When this first step is completed, press Next to view and edit layer 1. On that
layer, many of the constructions are invisible, so select each one that is supposed to
be invisible and then select the invisible color. The keyboard shortcut that sets the
selected primitive to be invisible is Ctrl-i. Since the first step of the proof concerns
two angle bisectors, select those and set them to the blinking color, and since they are
angle bisectors, it might also be a good idea to select the angles themselves and make
them blink as well.
To complete this step, add text to describe this first step of the proof.
Continue in this way until all the steps are finished. In this presentation, it required
9 steps, ending with a view of layer 8. (Remember that the first step is on layer 0.)

11.3. BUILDING GEOMETER PROOFS AND CONSTRUCTIONS

251

It is now a good idea to go to the beginning (using the Start button) and step forward
and backward through the demonstration, checking to see that all the primitives that
were supposed to be invisible are, in fact, invisible. Correct any errors.
Next, edit the Geometer file with the Edit Geometery command in the Edit pulldown menu and add the line:
.l0;

at the beginning of the file so that when the diagram is loaded the viewer will see only
the contents of layer 0 (as if the Start button were pressed automatically upon opening
the file).
Finally, it is nice to add a line at the end of the file like this:
.text("Press Next to continue ...", [8 .red, .in]);

This line appears after everything, but appears in red on the first 8 layers. Thus it
will be displayed underneath the text you entered earlier on every step except the final
one. It helps the user to know when the end of the proof is reached.
Now, load the diagram Presenting/ThreeIns.T and step through it, going forward
and backward with the Next and Prev keys to see what a reasonably good proof looks
like. You might also find it educational to view the code with the Edit Geometry command. Finally, it would be very useful to try to reproduce this file, or something similar,
starting from scratch. You can then refer to Presenting/ThreeIns.T if you get stuck.

11.3.2 Squaring a Rectangle

We will finish this section with one more example, the presentation of a geometric
construction: Given an arbitrary rectangle, construct a square that has the same area.
The same techniques and cautions that were mentioned in Section 11.3.1 apply
here.
We first draw a Geometer diagram that does what we want and test it. Then we
will begin with the entire diagram, complete with all the construction lines, and convert
it to a step by step construction.
In this case, the solution is not too hard.If the dimensions of the rectangle are a
and b, then the side of the desired square is ab. If a circle has diameter a + b, then
the distance to thecircle along a line perpendicular to the diameter that splits it into
lengths a and b is ab.
The complete construction is illustrated in figure 11.2. The user can change the
shape of rectangle OP RQ by moving points Q or P . Those points are constrained to
be on vertical and horizontal lines emanating from O. Since OP extends to the right,
we can make a segment of length OP + OQ by copying OQ down to the line OP with
a circle centered at O passing through Q and S. Clearly SP = OP + OQ.
In the figure, T is the midpoint of SP , so a circle centered at T will have SP as
diameter. A perpendicular to SP at O will intersect the circle with diameter SP at U .

CHAPTER 11. GEOMETRIC PRESENTATIONS

252

Figure 11.2: Rectangle to Square

Presenting/RectSquare.T [P]

The required square has a side equal to AU , and a circle centered at A passing through
V will determine the two sides of the square. Finally, polygons with different colors
and stippling patterns are draw around the rectangle and the square so the student can
see their relative sizes and shapes.
Now we do much the same thing we did with the proof in Section 11.3.1. We press
the Start button and modify the layers so that each presents the next step. As before,
it is probably a good idea to show the adjustable rectangle and square in the initial
layer so the user can experiment and see what the proposed construction is supposed to
accomplish.
Probably the best way to display this first step is with only points P and Q visible,
as well as the two rectangles. Everything else is invisible.
On the next step (layer 1), eliminate everything but the magenta rectangle with
points O, P and Q visible. Also include the line OP and the circle centered at O of
radius OQ intersecting the line at S.
In the next step we introduce the midpoint of SP and notice that it was named T in
the original diagram. It would better perhaps to change the name to M to make it easier
to remember that it is the midpoint. Again, add a textual description. Since viewers
are convinced already that OQ = OS we can probably make the circle centered at O
passing through Q invisible.
Continue in this way, and when you are finished, as before, add commands with
the text editor to display only layer 0 when the file is loaded and another text entry
at the end that indicates to the user that pressing the Next key will present additional
information.
As before, load the file Presenting/RectSquare.T and see how it looks and works.
View the constructions with the text editor to see in what order the primitives were
added. Make sure you understand how you could build it yourself.

11.4 Some Final Touches

Sometimes Geometer diagrams can be improved, either by making them less cluttered
or more robust. The next sections show a few examples where this is done.

11.4. SOME FINAL TOUCHES

253

11.4.1 Avoiding Jumping Points

Sometimes a Geometer diagram can be made more robust by using one of the commands LCP=>P Other or CCP=>P Other, both available under the Point submenu of
the Primitives pulldown menu.
A line and a circle or two circles generally intersect in two points or zero points.
Thus, when you use the Geometer command to construct a point that is at one of those
sorts of intersections, it solves the equations and finds two solutions. One of them is
arbitrarily called solution 1 and the other, 2 (depending of the sign of the square root
taken in the solution of a quadratic equation, if you know what that means). When the
graphical user interface is used to identify an intersection, Geometer uses the solution
closest to the final click of the mouse when the final circle is identified.
Unfortunately, this is only guaranteed to be correct at the time of creation.
As you move free points later, the intersection may jump suddenly to the other
possibility. Somewhat surprisingly, this
occurs rarely, or when it does occur, it
is masked because both intersections are
used. But the two commands above can
be useful for making this occurrence even
more rare.

D
B

Quite often you have a point that is

at one of the intersections and you would Figure 11.3: Angle Bisector Illustration
Presenting/Bisect.T [P]
like to obtain the other intersection. The
two commands above work exactly like
LC=>P and CC=>P except that they require a third parameterthe intersection point
you do not want to be chosen. When the more complex commands are executed, Geometer finds both intersections and chooses the one furthest from the point that is not
to be chosen.

11.4.2 Using Arcs

Especially in constructions, the entire circle is not needed, and if entire circles are
drawn, the figure can be quite cluttered. As an example, consider the construction of
an angle bisector.
The construction is illustrated in figure 11.3. Given ABC, construct a circle
centered at B and find its intersection at D with ray BC. Draw the two circles centered
at A and D that pass through B. They will have another intersection at E. (This is an
ideal place to use the CCP=>P Other discussed in Section 11.4.1.) The line BE is the
angle bisector of ABC.
The figure is certainly correct, but it is a bit cluttered. Compare it with figure 11.4
where small arcs of circles are used instead of the entire circles.

254

CHAPTER 11. GEOMETRIC PRESENTATIONS

An arc is specified somewhat like an

C
angle. If three points A, B and C are
D
E
given, point A is the beginning of the arc,
point B is the center of the circle upon
which the arc lies, and BC is the ray that
B
A
marks the end of the arc. The arc goes
counterclockwise from A to the ray BC.
Thus to specify an arc we need beginFigure 11.4: Better Angle Bisector Illusning and ending points for it that are intration
visible in figure 11.4. The way those were
Presenting/Bisect.T [P]
constructed is that two additional points
were placed in the diagram a short distance apart (which could be adjusted later). Using these two points as a radius, small
circles were drawn centered at points A, D and E. The intersections of these small
circles with the larger circles in figure 11.3 were determined, and those points were
used as the endpoints of the arcs.
If you cannot visualize this, look at the Geometer code for Presenting/Bisect.T
or load the diagram and press the Next key twice. The first press shows the intermediate
construction and the second shows the final state with the non-essential primitives set
to the invisible color.

11.5 Calculation in Geometer

Geometer has a built-in calculator that works with floating point numbers. The source
of the numbers can be from the distance between points, the measure of an angle, the
coordinates of points, the area of a polygon, a current value of a .script command
(see Section 11.7), or simply constants typed into a Geometer program.
These calculations can be useful directly for constructions, or for writing scripts
that allow Geometer to display an entire animation at the press of a button. See Section 11.7.
Perhaps the most useful way to use floating point numbers is simply to display them
on the screen. See Section 11.5.3.

11.5.1 Obtaining and Using Floating Point Numbers

Here is a list of commands to obtain floating point numbers for calculations:
You can simply enter them as 12 or 3.14159.
.f.vv: Generate a number that is the distance between two points.
.f.area: Find the (non-negative) area of a polygon and assign it to a number.
.f.vvvratio: Find the number that represents the ratio of the distance between
the first and second points to the distance between the first and third.

11.5. CALCULATION IN GEOMETER

255

.f.vxcoord: Assign the x-coordinate of a point to a number.

.f.vycoord: Assign the y-coordinate of a point to a number.
You can also use angles directly as floating point numbers. The angles are measured in degrees or radians, depending on the mode of Geometer. By default,
Geometer is in degree mode.
Once you have some numbers to work with, you can do arbitrary calculations with
the command .f.rpn (discussed below).
After you have done your calculations, you can use the results of your calculations
as follows:
.v.ff: Create a point whose coordinates are given by two floating point numbers.
.c.vf: Create a circle with a point as center and number as radius.
.a.f: Create an angle whose value is the floating point number. The angles size
is set depending on whether Geometer is in degree or radian mode.
Floating point numbers can be used in almost all of the transformation commands. See Section 11.6.

11.5.2 Calculating with Floating Point Numbers

Finally, you need to know how to calculate. All calculations are done with a single
function, .f.rpn. The rpn stands for reverse Polish notation which is a stack-based
system of calculation that eliminates the need for parentheses and complex parsing.
In the case of Geometer, the stack is simply an initially empty pile of numbers,
and each new number is conceptually piled on top of the others. You can only access
the stack from the topsort of like a pile of trays in a cafeteria. You can put trays on
top or take them off the top, but it is difficult to access the trays below without first
removing the ones on top.
The .f.rpn command is a list of numbers and operations that are evaluated from
left to right. If it is a number or a variable that contains a number, the value of that
number is simply put on the top of the stack. If it is an operation, that operation is
applied to the top few elements of the stack and the result is usually put back on top of
the stack. At the end of the evaluation, whatever number is left on top of the stack is
the result of the calculation.
For example, the operation .add adds the top two stack elements, removing them
from the stack, and the sum is put back on the stack. The following command assigns
the value of 5 to r:
r = .f.rpn(2, 3, .add);

CHAPTER 11. GEOMETRIC PRESENTATIONS

256

The first two items are numbers, so each is put on top of the stack. When it is
time to evaluate the .add, the stack consists of a 3 on top of a 2. The .add command
removes the top two numbers (all of them in this case), adds them to obtain 5, and then
places the 5 on top of the stack. Since the .add is the final operation in this example,
the value of 5 is assigned to r.
An application can be as simple as this:
p = .f.rpn(0.95);

which simply assigns the value 0.95 to the variable r.

It takes a bit of practice, but it is possible to evaluate extremely complex functions
using this simple notation. For example, suppose we would like to evaluate:
r = sin(x2 3x) cos(1/x).
Here is the statement that will accomplish it:
r = .f.rpn(x, x, .mul, 3, x, .mul,
.sub, .sin, 1, x, .div, .cos, .mul);

The function puts on two copies of x, multiplies them, and returns the value of x 2
to the stack. Then a 3 and an x are added on top of the x2 . Those top two numbers are
multiplied and replaced on the stack leaving an x2 under a 3x. Those two values are
subtracted and the result is returned to the stack, leaving x2 3x on the stack. Next, the
sine function is applied to the top stack element and the result is returned to the stack
leaving a sin(x2 3x). The process continues in this way until the entire expression
is evaluated. Geometer understands about 30 operations and all are documented in the
reference manual.
In the sections that follow, a few additional examples appear, but as one final example, here is some code to calculate the area of a triangle given the lengths of the sides
using Herons formula. If a, b, and c are the lengths of the sides of a triangle, then if
s = (a + b + c)/2 is the semiperimeter, Herons formula tells us that the area of the
triangle is given by:
p
A = s(s a)(s b)(s c).

Here is Geometer code to do the calculation, supposing that the points are internally
named v1, v2 and v3:
a
b
c
s
A

=
=
=
=
=

.f.vv(v1,
.f.vv(v2,
.f.vv(v3,
.f.rpn(a,
.f.rpn(s,
.mul, s,

v2);
v3);
v1);
b, c, .add, .add, 2, .div);
s, a, .sub, .mul, s, b, .sub,
c, .sub, .mul, .sqrt);

11.5. CALCULATION IN GEOMETER

257

11.5.3 Displaying Floating Point Numbers

If a floating point number has a name, it, together with its name, is displayed on the
Geometer screen.
In the calculation in the previous section we did not display any of the information
on the screen. If the purpose of this demonstration were to present a triangle that the
user could manipulate and have the lengths of the sides as well as the area displayed, the
following Geometer code would do the trick. This code can be found in the Geometer
diagram Presenting/Triarea.T.
.geometry "version 0.60";
v1 = .free(-0.323353, 0.00898204, "A");
v2 = .free(0.161677, 0.461078, "B");
v3 = .free(0.338323, -0.161677, "C");
l1 = .l.vv(v1, v2);
l2 = .l.vv(v2, v3);
l3 = .l.vv(v3, v1);
c = .f.vv(v1, v2, "AB");
a = .f.vv(v2, v3, "BC");
b = .f.vv(v3, v1, "CA");
s = .f.rpn(a, b, c, .add, .add,
2.000000, .div);
A = .f.rpn(s, s, a, .sub, .mul,
s, b, .sub, .mul, s,
c, .sub, .mul, .sqrt, "A(\triangleABC)");

This code would generate a continuously updated display in the Geometer window
that might look something like this:
AB = 0.882
BC = 0.547
CA = 0.683
A(4ABC) = 0.187
As we recall, the Geometer coordinate system goes roughly from 1.0 to 1.0 in
both dimensions. Students can probably think more clearly about numbers that are in a
more friendly rangesay from 1 to 10. The diagram above can easily be modified so
that the sizes of the legs of the triangle are roughly in this range by multiplying all the
lengths of the sides by a constant before applying Herons formula. If we want the sides
to have maximum lengths on the order of 10 to 15, then a multiplication by 5 will do
the trick. This code can be found in the Geometer diagram Presenting/Triarea5.T.
.geometry "version 0.60";
v1 = .free(-0.269461, 0.0299401, "A");
v2 = .free(-0.0898204, 0.601796, "B");
v3 = .free(0.682635, 0.39521, "C");
l1 = .l.vv(v1, v2);
l2 = .l.vv(v2, v3);
l3 = .l.vv(v3, v1);
cg = .f.vv(v1, v2);
ag = .f.vv(v2, v3);
bg = .f.vv(v3, v1);

258
c
a
b
s

CHAPTER 11. GEOMETRIC PRESENTATIONS

=
=
=
=

.f.rpn(cg, 5.000000, .mul, "AB");

.f.rpn(ag, 5.000000, .mul, "BC");
.f.rpn(bg, 5.000000, .mul, "CA");
.f.rpn(a, b, c, .add, .add,
2.000000, .div);
A = .f.rpn(s, s, a, .sub, .mul,
s, b, .sub, .mul, s,
c, .sub, .mul, .sqrt, "A(\triangleABC)");

In this example it might even be better to display a coordinate system to help the
students visualize the size of the triangles legs. The coordinate system could be drawn
with .pinned points so the user could not move it. Of course if Geometers coordinates
are all multiplied by 5 as in the last example, so would they be in the coordinate system.
Here is some code that does just that found in Presenting/Coordsys.T.
.geometry "version 0.60";
v0 = .pinned(-0.6, 0, .plus, "-3");
v1 = .pinned(-0.4, 0, .plus, "-2");
v2 = .pinned(-0.2, 0, .plus, "-1");
v3 = .pinned(0, 0, .plus, "0");
v4 = .pinned(0.2, 0, .plus, "1");
v5 = .pinned(0.4, 0, .plus, "2");
v6 = .pinned(0.6, 0, .plus, "3");
v7 = .pinned(0, -0.6, .plus, "-3");
v8 = .pinned(0, -0.4, .plus, "-2");
v9 = .pinned(0, -0.2, .plus, "-1");
v10 = .pinned(0, 0, .plus, "0");
v11 = .pinned(0, 0.2, .plus, "1");
v12 = .pinned(0, 0.4, .plus, "2");
v13 = .pinned(0, 0.6, .plus, "3");
l1 = .l.vv(v0, v6, .longline);
l2 = .l.vv(v13, v7, .longline);

For a triangle area demonstration, it would be nice to combine the following code
with that in the example above. You cannot, however, just put the code together in one
file since there are name conflicts. There is a v1, v2, v3, l1 and l2 in both files. If you
have something like the coordinate system above that you would like to use in multiple
examples you have two options. First, use very unusual primitive names in the file that
you will insert into the other files, or second, use the Geometer Insert command. If
you load the file with the triangle area demonstration into Geometer and then use the
Insert command to add the Presenting/Coordsys.T file you will find that Geometer
modified all the names in the inserted file.
The Geometer commands look something like this after the insertion:
v0__b38f9
v1__b38f9
v2__b38f9
...
l1__b38f9
l2__b38f9

= .pinned(-0.6, 0, .plus, "-3");

= .pinned(-0.4, 0, .plus, "-2");
= .pinned(-0.2, 0, .plus, "-1");
= .l.vv(v0__b38f9, v6__b38f9, .longline);
= .l.vv(v13__b38f9, v7__b38f9, .longline);

11.6. USING GEOMETRIC TRANSFORMATIONS

259

11.6 Using Geometric Transformations

Although they can certainly be used alone, the transformation primitives are useful in
Geometer primarily for constructing scripts. If you want to have a point move across
the screen or in a circle, or in a more complicated path, often the easiest way to do this
is with transformations.
In this section we will examine transformations used alone, but once you understand how to do this and how simple scripts work, it is not too difficult to use the transformations in Geometer scripts. See Section 11.7.3 for a non-trivial application of
transformations, and look at the source code for the circle demonstration in figure A.26
for a vastly more complicated exampleso complicated, in fact, that a computer program was used to generate it.
A Geometer transformation is a rule that tells how to move a point. It is possible
to use any linear transformation within Geometer, but the most common ones are
translation, rotation, scaling and combinations of those.
Every transformation is simply a modification of another transformation with the
exception of the identity transformation that does nothing. To build a Geometer transformation, you begin with the identity transformation and modify it until it does what
you want.
We begin with a very simple example. It will contain one free point A, and two
translations of A. See Presenting/Trans1.T. Here is the entire code:
v1
id
t1
t2
v2
v3

=
=
=
=
=
=

.free(-0.227545, 0.110778, "A");

.x.identity();
.x.translate(id, 0.400000, 0.000000);
.x.translate(id, -0.200000, 0.300000);
.v.vx(v1, t1, "A+(.4, 0)");
.v.vx(v1, t2, "A+(-.2, .3)");

Two different translations are built from the identity. The transformation t1 translates points by 0.4 units to the right and t2 translates them 0.2 units to the left (which
is the negative of right) and 0.3 units up. Point v2 is the result of applying t1 to v1 and
v3 is the result of applying t2 to v1. Load the diagram and move point A. The other
two stay at those fixed distances from A.
As a second example, the file Presenting/Trans2.T rotates a triangle by 75 about
the origin. Here is the code:
v0 = .pinned(0, 0, "O");
v1 = .free(0.311377, 0.11976, "A");
v2 = .free(0.58982, 0.41018, "B");
v3 = .free(0.610778, 0.0718563, "C");
l1 = .l.vv(v1, v2);
l2 = .l.vv(v2, v3);
l3 = .l.vv(v3, v1);
id = .x.identity();
rot = .x.rotate(id, 75.000000);
w1 = .v.vx(v1, rot, .green, "A");

CHAPTER 11. GEOMETRIC PRESENTATIONS

260
w2
w3
l4
l5
l6

=
=
=
=
=

.v.vx(v2,
.v.vx(v3,
.l.vv(w1,
.l.vv(w2,
.l.vv(w3,

rot, .green, "B");

rot, .green, "C");
w2, .green);
w3, .green);
w1, .green);

The file begins with a point O pinned at the origin (0, 0). All rotation occurs about
the origin, so the original triangle 4ABC will be rotated about that point by 75 . The
next six lines of code that displays 4ABC were drawn with Geometers graphical
user interface. The next two lines build a 75 rotation beginning with the identity and
combining a rotation with it. Finally, the points w1, w2 and w3 are rotated versions of
the vertices of the original triangle. Load the diagram and modify points A, B and C
to see how it works.
Finally, let us modify the example above to do a rotation about an arbitrary point
P . This will be done by translating P to the origin, doing a rotation about the origin,
and finally moving P back out to its original location. Some arithmetic must be done
to achieve this.
Here is the complete code for this more complex transformation:
vp = .free(-0.0958084, -0.239521, "P");
v1 = .free(0.248503, 0.00598802, "A");
v2 = .free(0.667665, 0.293413, "B");
v3 = .free(0.610778, 0.0718563, "C");
l1 = .l.vv(v1, v2);
l2 = .l.vv(v2, v3);
l3 = .l.vv(v3, v1);
dx = .f.vxcoord(vp);
dy = .f.vycoord(vp);
dxneg = .f.rpn(dx, -1.000000, .mul);
dyneg = .f.rpn(dy, -1.000000, .mul);
id = .x.identity();
xform1 = .x.translate(id, dxneg, dyneg);
xform2 = .x.rotate(xform1, 75.000000);
rot = .x.translate(xform2, dx, dy);
w1 = .v.vx(v1, rot, .green, "A");
w2 = .v.vx(v2, rot, .green, "B");
w3 = .v.vx(v3, rot, .green, "C");
l4 = .l.vv(w1, w2, .green);
l5 = .l.vv(w2, w3, .green);
l6 = .l.vv(w3, w1, .green);

This time the point P is the center of rotation and it is free so can be moved anywhere in the window. the variables dx and dy are the distances of the point P from
the origin. Calculations are done to produce the negatives of these values: dxneg and
dyneg. Next, after the identity transform is produced, it is successively modified to
translate it to the origin, to do a 75 rotation, and then the translate back to the original
point. The Geometer diagram is found in Presenting/Trans3.T. It seems similar to
the previous example except that in this case, the center of rotation P can be moved.

11.7. CONSTRUCTING GEOMETER SCRIPTS

261

11.7 Constructing Geometer Scripts

A script is based on a very simple idea. A floating point variable is successively
set to different values at a uniform rate of speed. For example, a script could simply set the value of the variable t to values between 0.0 and 1.0 in steps of 0.01.
This script would redraw the geometer diagram 101 times with t taking on the values
0.00, 0.01, 0.02, 0.03, . . ., 0.99, 1.00.
Assuming that the value of t is used in the construction of the diagram, each step
in the script will draw a different version.
As a very simple example, here is a Geometer diagram that will move a point
along the x-axis in 101 steps from x = 1.0 to x = 1.0. This example can be found
in Geometer diagram Presenting/Script1.T.
t = .script(-1.000000, 1.000000, 0.020000);
v0 = .v.ff(t, 0.000000);

The first line defines the script variable t to begin at 1.0 and to end at 1.0 taking
steps of 0.02 between each rendering of the scene. If you run this script by clicking on
the Run Script in the command area, a point moves across the center of the drawing
area in uniform steps.
An easy modification to the file is to change the second line to:
v0 = .v.ff(t, 0.000000, .smear, .dot);

This changes the style of the point to be drawn as a single dot, and to change the
color to the smearing color. When this script is run, the point steps uniformly across
the screen, but only dots are drawn, and they are not erased after each rendering, so a
line of dots is drawn. This modification is in the file Presenting/Script2.T.
Another small modification (available in file Presenting/Script3.T) changes the
diagram into something that will plot the curve y = x2 for 1.0 x 1.0:
t = .script(-1.000000, 1.000000, 0.020000);
t2 = .f.rpn(t, t, .mul);
v0 = .v.ff(t, t2, .smear, .dot);

The second line calculates the value of t2 by multiplying t by itself. The plotted
point has coordinates (t, t2 ) and is drawn in the smearing color.
That is all there is to it! A Geometer diagram can contain only a single script variable, and everything that changes has to be based on different values of that variable.
Obviously things are going to be much more interesting if we can somehow control
some geometric shapes and positions with a script variable. In addition to the floating
point calculator commands, the commands that tend to be immediately valuable with
the script variable are:

CHAPTER 11. GEOMETRIC PRESENTATIONS

262

.v.vvf: Produces a point between two others a given ratio of the distance between them. If the floating point value is zero, the new point lies on the first one.
If it is one, it lies on the second. A value of 0.5 puts it midway between them.
Values less than zero or greater than one also make sense.
.v.ff: Produce a point with coordinates given by two floating point values.
.c.vf: Produce a circle centered at a point with a radius given by the floating
point value.
.a.f: Produce an angle equal to the floating point value. If the diagram is in
degree mode (the default), a value of 90 makes a right angle. If the diagram is in
radian mode, a right angle corresponds to a floating point value of /2.
Any of the transformation commands. See Section 11.6.

11.7.1 The Witch of Agnesi

Perhaps the most common use of scripts in Geometer is to illustrate the solution to a
locus problem, of which the witch of Agnesi is one.

B
X

A
P

Figure 11.5: The Witch of Agnesi

Presenting/Witch.T [M]

Given a circle and a horizontal line l tangent to that circle, let P be the point on the
circle diametrically opposite the point of tangency with the line. Any secant though P
passes through the circle at a point A and through the line l at a point B. The witch of
Agnesi is the locus of points X that lie on the intersection of the vertical line through
B and the horizontal line through A. Figure 11.5 shows a portion of the curve.
Figure 11.5 is certainly enough to understand roughly what is going on, but if we
wish to make a nicer version with more evenly spaced points, we can do so with a
script.
Here is the code for Presenting/Witch.T:
.geometry "version 0.60";
v1 = .pinned(-1.11814, 0.649789, .in);
v2 = .pinned(1.09705, 0.649789, .in);
l1 = .l.vv(v1, v2, .longline);
v3 = .vonl(l1, 0, 0.649789, .in);
l2 = .l.vlperp(v3, l1, .in, .longline);

11.7. CONSTRUCTING GEOMETER SCRIPTS

v4
c1
v5
v6
l3
v7
l4
l5
v8

=
=
=
=
=
=
=
=
=

263

.vonl(l2, 0, 0.0442478, .in);

.c.vv(v4, v3);
.v.lcvother(l2, c1, v3, "P");
.vonl(l1, -1.84971, 0.649789, "B");
.l.vv(v5, v6, .longline);
.v.lcvother(l3, c1, v5, "A");
.l.vlperp(v6, l1, .longline);
.l.vlpar(v7, l1, .longline);
.v.ll(l4, l5, .smear, .dot, "X");

The point B is dragged along the line, and appropriate intersections are found that
eventually result in the point X which moves along the witch.

Figure 11.6: The Witch of Agnesi

Presenting/ScriptWitch.T [S]

To convert this to a script, all we need to do is make B move along the horizontal
line uniformly. This could be accomplished by replacing the line:
v6 = .vonl(l1, -1.84971, 0.649789, "B");

with this:
t = .script(-4.0, 4.0, .05);
v6 = .v.ff(t, 0.649789, "B");

but this has the slight disadvantage that if the figure is scaled, only the y-coordinates of
points will change; not the hard-coded value 0.649789 in the proposed solution.
Here is a better way: replace that same line defining point B with the following:
t = .script(-4.000000, 4.000000, 0.050000);
vx = .v.ff(t, 0.000000);
lx = .l.vlperp(vx, l1);
v6 = .v.ll(lx, l1, "B");

The difference is that a point vx is created that moves horizontally in a smooth

manner. A perpendicular from vx is dropped to the line l1 and the intersection of l1
and that perpendicular line lx identify the point B. You can test this version which can
be found in the file Presenting/ScriptWitch.T. The result is illustrated in figure 11.6.

264

CHAPTER 11. GEOMETRIC PRESENTATIONS

11.7.2 Poncelets Theorem Demonstration

In Section 3.8 we constructed a diagram to illustrate Poncelets theorem for a quadrilateral simultaneously tangent to an inner circle and inscribed in an outer one. In the
demonstration in that chapter the user can move the point O, the center of the inner
circle, and then leaving the point O in place move the point A around the outer circle
and verify that for any starting position the quadrilateral remains tangent to the inner
and inscribed in the outer circle.
A trivial modification changes that demonstration to a script that allows the user to
adjust the position of O as before, but then the Run Script command drives the point
A around the circle making two complete loops before it stops. The new code (in file
Presenting/Script4.T) looks like this:
R
O
C
V
d
r

=
=
=
=
=
=

.f.rpn(0.950000);
.v.ff(0.000000, 0.000000, .in);
.c.vf(O, R);
.free(-0.125749, 0.191617, "O");
.f.vv(O, V);
.f.rpn(1.000000, R, d, .add, .dup,
.mul, .div, 1.000000, R, d,
.sub, .dup, .mul, .div, .add,
1.000000, .exch, .div, .sqrt);
c = .c.vf(V, r);
//v1 = .vonc(C, 0.884919, 0.34557, "A");
t = .script(0.000000, 720.000000, 2.000000);
x = .f.rpn(t, .cos, R, .mul);
y = .f.rpn(t, .sin, R, .mul);
v1 = .v.ff(x, y, "A");
l1 = .l.vc(v1, c, 2);
v2 = .v.lcvother(l1, C, v1, "B");
l2 = .l.vc(v2, c, 2);
v3 = .v.lcvother(l2, C, v2, "C");
l3 = .l.vc(v3, c, 2);
v4 = .v.lcvother(l3, C, v3, "D");
l4 = .l.vv(v4, v1);
l5 = .l.vv(v1, v2);
l6 = .l.vv(v2, v3);
l7 = .l.vv(v3, v4);

The only modification from the old code is that the original line that defined v1 as
a point constrained to lie on a circle has been commented out, and replaced by the four
lines that follow it. (The // at the beginning of the line containing the old definition
of v1 indicates that the rest of the line is commented out and is invisible to Geometer.)
The first of those lines defines a script that causes the value of t to vary from 0 to
720 (degrees). The following two lines calculate the location of the point on the circle
by taking the sine and cosine of t and multiplying the results by the radius R. Finally,
a new version of v1 is created with x and y as its coordinates. All the rest of the code,
before and after, is identical.

11.7. CONSTRUCTING GEOMETER SCRIPTS

b

a
b

265

c
b

Figure 11.7: Proof 2 of the pythagorean theorem

Review/Pyproof2.D, Review/Pyproof3.D [M]

11.7.3 Script Plus Transformations

It is possible to construct animations with Geometer using the transformation commands discussed in Section 11.6. In this section we will construct a dynamic version
of the pythagorean theorem demonstrated by the illustration shown in Section A.6.5,
reproduced here as figure 11.7.
We would like to have the four triangles in the figure begin in one position and the
Run Script button is pressed, to have them move to form the other configuration. A
quick look at figure 11.7 shows us there is a problem: if we use the triangles as they
appear in that figure, two of them will need to be flipped over during the transformation.
In other words, in the diagram on the left, all the triangles are right-handed and in the
diagram on the right, two are right-handed and two are left-handed. If you cannot
see this, try cutting out the triangles from small pieces of paper and see what happens
if you try to slide them around. Eventually you will need to invert two of them.
A

B
A

C
C

Figure 11.8: Dynamic Pythagorean Theorem

Presenting/Script5.T [S]

A better pair of initial and final configurations are shown in figure 11.8. In this
section we will describe how to build a Geometer script that moves smoothly from
the configuration on the left to that on the right. Notice that all four triangles in this
diagram have the same handedness.
We would like to allow the user to adjust the shape of the triangle before running
the script, but we know that the four triangles will initially have to be arranged around a
square as in the diagram on the left of figure 11.8. The easiest way to do that is to fix the

266

CHAPTER 11. GEOMETRIC PRESENTATIONS

square and to allow the point B to move along the edge. Because of the arrangement of
the triangles, the location of B will determine the length AC as well, and all possible
right triangle shapes can be illustrated with this single degree of freedom.
In the initial diagram it is important to analyze the motions that the four triangles
must make. If we number the initial ones 1, 2, 3 and 4 beginning with the one labeled
4ABC in the upper left and continuing clockwise, triangle 1 needs to be translated
straight down by a distance equal to AB. Triangle 2 does not move, triangle 3 rotates
counterclockwise for 90 about its uppermost tip, and triangle 4 rotates 90 clockwise
as its uppermost point slides down to the lower left corner of the surrounding square.
The easiest way to build the script is one transformation at a time. Write each of
the three transformations and test it before moving on to the next. Remember that
one of the triangles does not need to move at all so there are three, rather than four
transformations.
Here is the entire code for Presenting/Script5.T:
.geometry "version 0.60";
t = .script(0.000000, 1.000000, 0.010000);
v0 = .pinned(-0.85, 0.85);
v1 = .pinned(-0.85, -0.85);
v2 = .pinned(0.85, 0.85);
l1 = .l.vv(v0, v2);
l2 = .l.vv(v0, v1);
v3 = .vonl(l1, -0.320359, 0.85);
c3 = .c.ctrvv(v0, v2, v3, .in);
v5 = .v.lc(l2, c3, 2);
l4 = .l.vlpar(v1, l1, .in);
l5 = .l.vlpar(v2, l2, .in);
v6 = .v.ll(l5, l4);
l6 = .l.vv(v2, v6);
l7 = .l.vv(v6, v1);
c4 = .c.ctrvv(v1, v0, v5, .in);
c5 = .c.ctrvv(v6, v0, v5, .in);
v7 = .v.lc(l6, c5, 1);
v8 = .v.lc(l7, c4, 1);
ab = .f.vv(v0, v3);
xid = .x.identity();
dy = .f.rpn(t, ab, .mul, -1.000000, .mul);
xabc = .x.translate(xid, 0.000000, dy);
a1 = .v.vx(v0, xabc, "A");
a2 = .v.vx(v3, xabc, "B");
a3 = .v.vx(v5, xabc, "C");
poly1 = .polygon(3, a1, a2, a3, .red, .hashpoly);
ex = .f.vxcoord(v7);
ey = .f.vycoord(v7);
nex = .f.rpn(ex, -1.000000, .mul);
ney = .f.rpn(ey, -1.000000, .mul);
xe1 = .x.translate(xid, nex, ney);
xang = .f.rpn(t, -90.000000, .mul);
xe2 = .x.rotate(xe1, xang);
xe3 = .x.translate(xe2, ex, ey);
b1 = .v.vx(v7, xe3);
b2 = .v.vx(v6, xe3);
b3 = .v.vx(v8, xe3);

11.7. CONSTRUCTING GEOMETER SCRIPTS

267

poly2 = .polygon(3, b1, b2, b3, .green, .hashpoly);

poly3 = .polygon(3, v2, v3, v7, .yellow, .hashpoly);
cx = .f.vxcoord(v5);
cy = .f.vycoord(v5);
ncx = .f.rpn(cx, -1.000000, .mul);
ncy = .f.rpn(cy, -1.000000, .mul);
xc1 = .x.translate(xid, ncx, ncy);
yang = .f.rpn(t, 90.000000, .mul);
xc2 = .x.rotate(xc1, yang);
xc3 = .x.translate(xc2, 0.000000, dy);
xc4 = .x.translate(xc3, cx, cy);
d1 = .v.vx(v1, xc4);
d2 = .v.vx(v5, xc4);
d3 = .v.vx(v8, xc4);
poly4 = .polygon(3, d1, d2, d3, .cyan, .hashpoly);

The first line defines a script variable t that varies from 0.0 to 1.0 in steps of 0.01.
The 0.01 is not too importantwe can adjust it later if the script seems to run too slow
or too fast. The range 0.0 to 1.0 is arbitrarywe will have to transform it to move
angles from 0 to 90 and to translation amounts that depend on the lengths of the
sides of 4ABC.

The next few lines draw the surrounding square and define a point B that is constrained to move on the upper line of the square. Notice that the point constrained to
be on the line does not have the name B. This is because the real point B will be a
transformation of that point. Initially the transformation will be zero, but although the
point v3 remains on the top edge of the enclosing square, its transformed image called
B will move down as the script proceeds.
In fact, in the construction of the diagram, all the vertices defining the four triangles
were visible and had readable names to make it easier to find the appropriate points
to transform. When everything was working, the names were eliminated and certain
constructions were made invisible.
The next lines are standard Geometer constructions of the rest of the vertices on the
square whose positions vary based on the location of the controlling point v3 (which
looks like it is named B, but is not).
The script work begins with the line:
ab = .f.vv(v0, v3);

This line assigns the length of the segment AB to a variable called ab. Next we
begin building a transformation that will slowly slide the triangle in the upper left
downward. As all transformations must begin with an identity transform, that is what
is done in the next line. The amount of movement dy is calculated in the next line,
and it depends on the script variable t. As t varies from 0.0 to 1.0 dy varies from
0.0 to AB. Next, xabc is produced by a translate command that moves objects by a
distance dy in the y-direction and does not move them at all in the x-direction. Finally,
transformation xabc is applied to the three initial vertices of the triangle in the upper
left corner to produce the version with vertices labeled A, B and C that slides down
the window.

268

CHAPTER 11. GEOMETRIC PRESENTATIONS

A polygon is drawn connecting these three vertices so that the solid effect will be
obvious.
Next, a similar procedure is written to rotate the triangle in the lower right for 90
about its upper point. We cannot simply use a rotation command, since the rotation
transformation rotates about the origin. What we must do is translate the point to the
origin, rotate it there, and then translate back to the original position. That is why
the transformation xe3 is built in a series of steps: start with the identity, translate
to the origin, do the rotation, and finally, translate back. When it is ready, the xe3
transformation is applied to the three vertices of the triangle in the lower right, and
those transformed points b1, b2 and b3 are connected to form a polygon.
The final transformation is almost the same, but it is a combination of a rotation
and a translation. Thus the point about which the rotation is to occur is translated to the
origin, it is rotated and translated, and finally translated back to its original position,
and the grand transformation that accomplishes all of this is called xc4. The original
vertices are transformed by xc4 and are also connected to form a polygon.

11.7.4 Scripts with Conditionals

The example in this section is not very interesting, but it illustrates a technique that
can be used in very interesting situations. Here we will construct a Geometer diagram which will move the center of a fixed-size circle around a square whose vertices
are (1/2, 1/2). For an illustration of a much more interesting example that was
constructed using similar techniques, look at Presenting/PythagProofFinal.T.
What we will do is to move the center first from (1/2, 1/2) uniformly to the
point (1/2, 1/2). Then we will move it from there to (1/2, 1/2), then to (1/2, 1/2)
and finally back to the start. If t varies from 0 to 1, a movement along the first path
can be achieved by taking a vertex having coordinates (1/2 + t, 1/2). Unfortunately,
when t reaches 1, a new formula is required that leaves the x-coordinate fixed and
moves along the y direction, and so on.
A technique that will work well is to have a single script where the variable t runs
from 0 to 4, and each time it advances by 1, a new segment of the square is traced.
Following is the code to achieve the desired result, and it can be found in the file

Presenting/SquareMove.T:

.geometry "version 0.62";

t = .script(0.000000, 4.000000, 0.040000);
in1 = .f.rpn(t, 1.000000, .lt);
in2 = .f.rpn(t, 1.000000, .ge, t, 2.000000,
.lt, .mul);
in3 = .f.rpn(t, 2.000000, .ge, t, 3.000000,
.lt, .mul);
in4 = .f.rpn(t, 3.000000, .ge);
t1 = .f.rpn(t);
t2 = .f.rpn(t, 1.000000, .sub);
t3 = .f.rpn(t, 2.000000, .sub);
t4 = .f.rpn(t, 3.000000, .sub);

11.7. CONSTRUCTING GEOMETER SCRIPTS

269

tx = .f.rpn(-0.500000, t1, .add, in1, .mul,

0.500000, in2, .mul, .add, 0.500000,
t3, .sub, in3, .mul, .add,
-0.500000, in4, .mul, .add);
ty = .f.rpn(-0.500000, in1, .mul, -0.500000, t2,
.add, in2, .mul, .add, 0.500000,
in3, .mul, .add, 0.500000, t4,
.sub, in4, .mul, .add);
v0 = .v.ff(tx, ty);
c0 = .c.vf(v0, 0.200000, .smear);

Figure 11.9: Circle Moving in a Square

The first line is the standard script running t from 0 to 4. The next four lines are
essentially boolean values: 1 (true) if t is in the range and 0 (false) otherwise. Note
that multiplication of zeroes and ones is the same as the boolan and operation. To
see if t is between 1 and 2 we just multiply 1 < t by t 2, and so on.

The next four steps produce for convenience values of t that vary from zero to one
in each of the ranges. Notice that we do not care about bad values outside the ranges,
since we will be multiplying the bad values by zero anyway.
The somewhat complex calculations for the centers (tx , tx ) of the circle effectively
multiply the position of the point along each side of the square by a boolean value that
is one only during the proper part of the path. Mathematically, it looks something like
this, if pi is only true (or equal to 1) on the correct part of the path. Here is a formula
for the x-coordinate. ti varies from 0 to 1 on the ith part of the path:
(1/2 + t1 )p1 + (1/2)p2 + (1/2 t3 )p3 (1/2)p4 .
A slightly different, but similar, calculation is done to find the y-coordinate, and
the two are combined in the next-to-last step to identify the center of the circle that will
follow a path along the square.
Figure 11.9 shows the result.

270

CHAPTER 11. GEOMETRIC PRESENTATIONS

11.8 Scripts and Layer Conditionals

A nice touch in a diagram is to have the circles and lines be drawn not all at once, but
rather as if they were being drawn by hand: an arc that expands to become a circle, or
as a line segment that starts as a point and gradually lengthens to its final length.
Figure 11.10 shows the final result of a
demonstration of the method to bisect a line segment that is not too interesting. What is interesting is to watch it being drawn by Geometer. As
the script runs, both circles are drawn smoothly
as is the line connecting points C and D, and the
points C, D and the midpoint M appear one at a
time.

C
A

D
The script uses the .layercondition command which turns on a set of layers that depends
on the value of the script variable, t in this case.
Figure 11.10: Smooth Drawing
In the example below, when t is between 0.0 and
Presenting/SmoothArc.T [S]
1.0, layer 1 is displayed and so on. Using this
trick it is simply a matter of drawing the correct
figure or animation in each of the layers. Each animation is drawn in one layer, and then
after that animation is complete, the finished figure appears in the following layers.
In this example, the original problem is presented on layer 0, the animation of the
first circle appears on layer 1, and a finished version of the circle appears on layers 2
and beyond. There is no animation on some of the layers so the person viewing the
script can think for a second or two about what was just done.
Also note that a diagram such as this can be viewed manually simply by repeatedly
pressing the Next button.
Here is the code:
.geometry "version 0.64";
.l0;
t = .script(0.000000, 30.000000, 0.010000);
.layercondition(0.000000, t, 1.000000, 0x1);
.layercondition(1.000000, t, 2.000000, 0x2);
.layercondition(2.000000, t, 3.000000, 0x4);
.layercondition(3.000000, t, 4.000000, 0x8);
.layercondition(4.000000, t, 5.000000, 0x10);
.layercondition(5.000000, t, 6.000000, 0x20);
.layercondition(6.000000, t, 7.000000, 0x40);
.layercondition(7.000000, t, 8.000000, 0x80);
.layercondition(8.000000, t, 9.000000, 0x100);
.layercondition(9.000000, t, 10.000000, 0x200);
.layercondition(10.000000, t, 11.000000, 0x400);
v1 = .free(-0.11976, 0.161677, "A");
v2 = .free(0.194611, 0.0359281, "B");
l1 = .l.vv(v1, v2);
v1x = .f.vxcoord(v1);
v1y = .f.vycoord(v1);

11.9. COMPUTER-GENERATED DIAGRAMS

271

v2x = .f.vxcoord(v2);
v2y = .f.vycoord(v2);
r12 = .f.vv(v1, v2);
theta1 = .f.rpn(v2x, v1x, .sub, v2y, v1y,
.sub, .exch, .atan2);
x = .f.rpn(t, t, .floor, .sub, 360.000000,
.mul, theta1, .add, .cos, r12,
.mul, v1x, .add);
y = .f.rpn(t, t, .floor, .sub, 360.000000,
.mul, theta1, .add, .sin, r12,
.mul, v1y, .add);
v3 = .v.ff(x, y, .in);
arc1 = .arc.vvv(v2, v1, v3, .l1);
circ1 = .c.vv(v1, v2, .l2on);
theta2 = .f.rpn(v1x, v2x, .sub, v1y, v2y,
.sub, .exch, .atan2);
x1 = .f.rpn(t, t, .floor, .sub, 360.000000,
.mul, theta2, .add, .cos, r12,
.mul, v2x, .add);
y1 = .f.rpn(t, t, .floor, .sub, 360.000000,
.mul, theta2, .add, .sin, r12,
.mul, v2y, .add);
v4 = .v.ff(x1, y1, .in);
arc2 = .arc.vvv(v1, v2, v4, .l3);
circ2 = .c.vv(v2, v1, .l4on);
p1 = .v.cc(circ1, circ2, 2, .l5on, "C");
p2 = .v.cc(circ1, circ2, 1, .l6on, "D");
frac = .f.rpn(t, t, .floor, .sub);
v5 = .v.vvf(p1, p2, frac, .in);
ltemp = .l.vv(p1, v5, .l7);
ll = .l.vv(p1, p2, .l8on);
mid = .v.vvmid(v1, v2, .l9on, "M");
.text("Construct the midpoint of the segment AB");
.text("Adjust the positions of A and B
and then press the Run Script button.", .red, .l0);

11.9 Computer-Generated Diagrams

Sometimes complex diagrams are required with hundreds or even thousands of lines.
When that occurs, it is inefficient to enter thousands of primitives using the Geometer
program. Geometer diagrams, however, are simply text files that can be generated by
any means, including computer programs.

11.9.1 The Koch Snowflake Curve

The first example generates the first few approximations of the Koch snowflake curve
one of the simplest examples of a fractal. The curve is illustrated in figure 11.11.
The first iteration is simply a triangle. The succeeding steps can be considered to be
successive modifications of the previous version.
To traverse the triangle, imagine following three straight-line paths. In each successive iteration, replace the movement along a path by a four-step path that follows

CHAPTER 11. GEOMETRIC PRESENTATIONS

272

the original path for 1/3 of its length, then turns to the right by 60 and continues for
the same distance. Next, turn 120 to the left, move another 1/3, and finally turn 60
to the right and follow the original curve to its end.
After each such modification, the curve is composed of straight-line segments; just
four times as many of them. The first six stages are illustrated in figure 11.11. When
the Geometer diagram is loaded, the triangle appears, but each time the Next button is
pressed, the next stage of the curve is presented.
The Geometer diagram simply consists of computer-generated sets of short line
segments that compose each stage, and those on each stage are presented on a new
layer.

Figure 11.11: The Koch Snowflake Curve

Presenting/Script5.T [S]

The computer code (written in the C language) appears below and on the CD as

koch.c:

#include <math.h>
#define PI 3.1415926535897
int vnum;
int glevel;
plot(double x, double y, double angle, double dist, int n)
{
double x1, y1;
if (n == 0) {
x1 = x + dist*cos(angle);
y1 = y + dist*sin(angle);
printf("v%dx%d = .v.ff(%f, %f, .nomark, .l%d);\n", glevel,

11.9. COMPUTER-GENERATED DIAGRAMS

273

vnum, x1, y1, glevel);

printf("l%dx%d = .l.vv(v%dx%d, v%dx%d, .l%d);\n", glevel, vnum, glevel,
vnum-1, glevel, vnum, glevel);
vnum++;
return;
}
plot(x, y, angle, dist/3, n-1);
x = x + dist*cos(angle)/3.0;
y = y + dist*sin(angle)/3.0;
angle -= PI/3;
plot(x, y, angle, dist/3, n-1);
x = x + dist*cos(angle)/3.0;
y = y + dist*sin(angle)/3.0;
angle += 2*PI/3;
plot(x, y, angle, dist/3, n-1);
x = x + dist*cos(angle)/3.0;
y = y + dist*sin(angle)/3.0;
angle -= PI/3;
plot(x, y, angle, dist/3, n-1);
return;
}
koch(int lev)
{
printf("v%dx0 = .v.ff(1.0, 0.0, .nomark, .l%d);\n", glevel, glevel);
plot(1.0, 0.0, 5*PI/6, 1.73205080757, lev);
plot(-0.5, .866025403785, 3*PI/2, 1.73205080757, lev);
plot(-0.5, -.866025403785, PI/6, 1.73205080757, lev);
}
main()
{
int i;
printf(".l0;\n");
printf(".text(\"The Koch (snowflake) curve\n\");");
printf(".text(\"Press n to continue ...\", .red, .tol3);\n");
for (i = 0; i < 5; i++) {
vnum = 1;
glevel = i;
koch(i);
}
}

Most of the work is done in the plot subroutine which is recursive and calls itself
for the correct number of levels, depending on the approximation to the Koch curve it
draws. If the level is zero, it simply draws a straight line of the given length in the given
direction and returns. If the level is not zero, it calls itself four times to make the four
subpaths, but with a level that is one less. Note that when each segment of the curve is
drawn, it is drawn on the same layer controlled by the variable glevel.
To draw the Koch curve, the plot routine is called three times by the koch routine
that draws paths that connect the three vertices of the original triangle.
Note: The actual Koch curve is none of the iterations aboveit is the limiting case
as the number of iterations increases without bound. It is interesting in that at each

CHAPTER 11. GEOMETRIC PRESENTATIONS

274

stage after the first one, the area increases by 2/9 of the amount added previously, but
the perimeter increases by a factor of 4/3. Thus the final, limiting Koch curve has finite
area (if the original triangle has area 1, then the area enclosed by the limiting curve is
10/7 = 1.42857 . . .) and infinite perimeter.

11.9.2 Apollonian Circle Packing

The next example is more complicated and it generates figure 11.12.
26

26

14

30

14

11

23

30
11

23

18

18

27

27
15

15
27

27
18

18
23

11
30

23
3

11
6

14

30

14
26

26

Figure 11.12: Apollonian Packing

Presenting/soddy.T [S]

The figure consists of a circle with two half-circles inside it. Beginning with those
three mutually-tangent circles, each set of three mutually-tangent circles will have circles tangent to those, and so on. If the holes are filled in forever, the result is called
the Apollonian gasket.

11.9. COMPUTER-GENERATED DIAGRAMS

275

The radii of successive circles can be calculated using the Descartes circle theorem
(see Section 2.14) and almost miraculously, if the radius of the outer circle is 1, then
all the radii of the inner circles are of the form 1/n, where n is some integer. In
figure 11.12 some of the larger circles have numbers inside them and those numbers
are the inverses of their radii. (For example, the circle with an 11 inside has radius
1/11 = .0909 . . ..) What is more, every circle with radius 1/n is centered at a point
with rational coordinates whose denominator is a divisor of n. With this information,
it is not too hard to write a computer program that generates the figure.
Here is the listing of such a program (called soddy.c on the CD) written in the C
language:
#include <math.h>
#include <stdlib.h>
#define RMAX 400
#define RPRINTMAX 30
// for point: center = (x1/x2, y1/y2); radius = 1/R
int vertnum = 5;
struct point {
int x1, x2, y1, y2, R;
};
int rightdist(struct point *p, int i, int j, int Rn)
{
int x1 = p->x1, x2 = p->x2, y1 = p->y1, y2 = p->y2, R = p->R;
if (R!=1 && (y2*y2*R*R*(x1*Rn-i*x2)*(x1*Rn-i*x2) +
x2*x2*R*R*(y1*Rn-j*y2)*(y1*Rn-j*y2) ==
x2*x2*y2*y2*(Rn+R)*(Rn+R))) return 1;
if (R==1 && (i*i+j*j == (Rn-1)*(Rn-1))) return 1;
return 0;
}
newcirc(struct point *p1, struct point *p2, struct point *p3, int depth)
{
int R1, R2, R3, Rnew;
int B, C, D, R;
int i, j;
struct point *p;
double Rad, Cx, Cy;
int imin, imax, jmin, jmax;
if
R1
R2
R3
if
if
if

(depth ==
= p1->R;
= p2->R;
= p3->R;
(R1 > R2)
(R2 > R3)
(R1 > R2)

0) return;

{ R = R1; R1 = R2; R2 = R; p = p1; p1 = p2; p2 = p;}

{ R = R2; R2 = R3; R3 = R; p = p2; p2 = p3; p3 = p;}
{ R = R1; R1 = R2; R2 = R; p = p1; p1 = p2; p2 = p;}

if (R1 == 1) R1 = -1;

276

CHAPTER 11. GEOMETRIC PRESENTATIONS

B = -2*(R1 + R2 + R3);
C = R1*R1 + R2*R2 + R3*R3 - 2*(R1*R2 + R2*R3 + R3*R1);
// now solve x^2 + Bx + C = 0;
D = B*B - 4*C;
R = (int)sqrt(D + .01);
Rnew = (R - B)/2;
if (Rnew < 0) Rnew = (-R- B)/2;
//printf("%d\n", Rnew);
Rad = 1.0/Rnew + 1.0/R3;
Cx = (double)(p3->x1)/p3->x2;
Cy = (double)(p3->y1)/p3->y2;
imin = (int)(Rnew*(Cx-Rad)-1);
imax = (int)(Rnew*(Cx+Rad)+1);
jmin = (int)(Rnew*(Cy-Rad)-1);
jmax = (int)(Rnew*(Cy+Rad)+1);
p = 0;
for (i = imin; i <= imax; i++)
for (j = jmin; j <= jmax; j++) {
if (rightdist(p1, i, j, Rnew) &&
rightdist(p2, i, j, Rnew) &&
rightdist(p3, i, j, Rnew)) {
p = (struct point *)malloc(sizeof(struct point));
p->x1 = i; p->x2 = Rnew;
p->y1 = j; p->y2 = Rnew;
p->R = Rnew;
if (Rnew < RMAX) {
newcirc(p1, p2, p, depth+1);
newcirc(p1, p, p3, depth+1);
newcirc(p, p2, p3, depth+1);
}
}
}
if (Rnew <= RPRINTMAX) {
printf("v%d = .pinned(%f, %f, .in);\n",
vertnum, (double)p->x1/Rnew, (double)p->y1/Rnew);
printf("vn%d = .v.vtranslate(v%d, -.037, -.037,"
".nomark, \"%d\", .l%don);\n",
vertnum, vertnum, Rnew, depth);
vertnum++;
}
if (Rnew < RMAX) {
printf("v%d = .pinned(%f, %f, .in);\n",
vertnum, (double)p->x1/Rnew, (double)p->y1/Rnew);
printf("c%d = .c.vf(v%d, %f, .l%don);\n", vertnum,
vertnum, 1.0/Rnew, depth);
vertnum++;
}
}
struct point O1, O2, O22, O3, O33;
main()
{
int n = 4;

11.9. COMPUTER-GENERATED DIAGRAMS

O1.x1 = O1.y1 = 0; O1.x2 = O1.y2 = 1; O1.R = 1;

O2.x1 = 1; O2.x2 = 2; O2.y1 = 0; O2.y2 = 1; O2.R
O22.x1 = -1; O22.x2 = 2; O22.y1 = 0; O22.y2 = 1;
O3.x1 = 0; O3.x2 = 1; O3.y1 = 2; O3.y2 = 3; O3.R
O33.x1 = 0; O33.x2 = 1; O33.y1 = -2; O33.y2 = 3;

277

= 2;
O22.R = 2;
= 3;
O33.R = 3;

printf(".l0;\n");
printf("v0 = .pinned(0,0, .in);\nc0 = .c.vf(v0, 1.0);\n");
printf("v1 = .pinned(0.5,0, .nomark, \"2\", .l1on);\n"
"c1 = .c.vf(v1, 0.5, .l1on);\n");
printf("v2 = .pinned(-0.5,0, .nomark, \"2\", .l1on);\n"
"c2 = .c.vf(v2, 0.5, .l1on);\n");
printf("v3 = .pinned(0,.66666666, .nomark, \"3\", .l2on);"
"\nc3 = .c.vf(v3, 0.3333333333, .l2on);\n");
printf("v4 = .pinned(0,-.666666666, .nomark, \"3\", .l2on);"
"\nc4 = .c.vf(v4, 0.3333333333, .l2on);\n");
newcirc(&O2, &O22, &O3, 3);
newcirc(&O2, &O22, &O33, 3);
newcirc(&O1, &O2, &O3, 3);
newcirc(&O1, &O22, &O3, 3);
newcirc(&O1, &O2, &O33, 3);
newcirc(&O1, &O22, &O33, 3);
}

Each circle has a point at its center rational coordinates, so the x- and y-coordinates
of each center point are listed with two integers: the numerator and denominator. A
radius (actually the inverse of the radius) is also part of the data structure.
Rather than calculate the coordinates of the center which requires some fairly messy
calculations, the program simply tries a small grid of possible centers. The routine
rightdist returns 1 if the given center point is the proper distance from a proposed
pair of rational coordinates. All calculations are done in integers, so exact results are
obtained in every case.
The routine newcirc generates the circle that lies in the gap between three circles,
and in addition, a parameter depth indicates how many levels down that circle is generated. It works by calculating the new radius and then checking that the center is the
appropriate distance from all three given circle centers. Finally, it calls itself recursively to create three more circles in the newly-created gaps. The recursion halts when
the circles get too small. In addition, numeric labels are printed on the diagram if the
circles are somewhat larger.
The main routine simply draws enough circles to get the recursion started (the first
five are necessary), and then newcirc is called on each triple of the initial circles. As
the circles are printed in the form of Geometer commands, the layer upon which each
appears depends on the depth at which that circle is generated. Thus the initial drawing
is only a large circle, but each time the Next command is issued, another level of
circles is drawn. In the sample code above, all circles of radius greater than 1/400 of
the original radius are drawn.

278

CHAPTER 11. GEOMETRIC PRESENTATIONS

Chapter

12

Geometer Proofs
This chapter contains a list of interesting theorems whose proofs are available on the
CD as Geometer files. Some of the theorems are those usually presented in high school
classes and could be used for classroom presentations by teachers. Others are proofs of
more complicated and generally more interesting theorems. In the text, each theorem
will be stated with an accompanying diagram. To see the proof, load the diagram into
Geometer and use the Next and Prev buttons to move forward and backward through
the steps.
To save space, unlike the rest of the book the diagrams here are untitled, and the
name of the Geometer file is listed with the problem description, not with the figure.
The more interesting and beautiful theorems are listed with an accompanying diagram; at the end of the chapter is a long list of problems without diagrams but for
which Geometer diagrams and proofs appear on the CD.
279

CHAPTER 12. GEOMETER PROOFS

280

12.1 Illustrated Theorems

12.1.1 Adams Circle
B

I
C

If 4ABC is any triangle, let A0 , B 0 and

C 0 be the points of tangency of the incircle with the sides of the triangle. Let
G (The Gergonne point) be the intersection of AA0 , BB 0 and CC 0 . The lines
through G parallel to the sides of the triC angle 4A0 B 0 C 0 intersect 4ABC at six
concentric points on Adams Circle.

J
H
M B

File: GeomProofs/Adams.T[P]

12.1.2 Rectangular Circle Centers

A
B

File: GeomProofs/Circlecenters.T[P]

If ABCD is a cyclic quadrilateral, show that

the incenters of the triangles 4ABC, 4BCD,
4CDA and 4DAB form a rectangle.

12.1.3 Fagnanos Problem

B

A
C

In any acute-angled triangle 4ABC, show

that if three points are selected inside the
segments AB, BC and CA the perimeter
of that triangle is greater than or equal to the
perimeter of the triangle formed when those
points are the feet of the altitudes. This minimizing triangle is called the orthic triangle. In the figure the orthic triangle is drawn
with solid black vertices.
File: GeomProofs/Fagnano.T[P]

12.1. ILLUSTRATED THEOREMS

281

12.1.4 The Euler Line

In any triangle 4ABC, show that the center of mass (Gthe intersection of the three
medians), the circumcenter (Othe center
of the circumscribed circle) and the orthocenter (Hthe intersection of the altitudes)
all lie on a line called the Euler line.

B
C

OG

H
A

File: GeomProofs/Euler.T[P]
File: GeomProofs/Euler1.T[P]

12.1.5 Similar Exterior Triangles

X

B
Y
A
C
Z

If three similar triangles are drawn exterior to the sides of 4ABC as shown
in the diagram on the left, show that the
circumcircles of all three similar triangles meet at a point.

File: GeomProofs/Simcircle.T[P]

12.1.6 The Fermat Point

E

C
D
O

Erect outward-pointing equilateral triangles

on all three sides of 4ABC. The lines connecting the vertices of 4ABC with the tips
of the equilateral triangles erected on the opposite sides meet at a point called the Fermat point.
File: GeomProofs/Fermat.T[P]

B
F

CHAPTER 12. GEOMETER PROOFS

282

12.1.7 Four Bisectors

K
L

Let ABCD be an arbitrary convex quadrilateral. If the interior angle bisectors are
constructed at A, B, C and D show that the
intersections of adjacent pairs of bisectors
are concyclic (all lie on the same circle).

File: GeomProofs/Fourbisectors.T[P]

12.1.8 Incenter Problem

A

O O
2

O1

If BAC = 90 in 4ABC, and AD is

the altitude from A to BC, show that the
line connecting the centers of the incircles
of 4ADB and 4ADC is perpendicular to
the segment connecting A to the incenter of
4ABC, and has the same length.
File: GeomProofs/IncenterDist.T[P]

12.1.9 Three Squares

Show that if three squares are arranged
as in the diagram then:

EBA + ECB + EDC = 90 .

A

File: GeomProofs/Squares.T[P]

12.1. ILLUSTRATED THEOREMS

283

12.1.10 Cyclic Quadrilateral Angle Bisectors

B

If ABCD is a cyclic quadrilateral and sides AB

and CD are extended to meet at E and BC and
AD are extended to meet at F , show that the angle bisectors of AED and BF A are perpendicular.

E
G

C
H
D

File: GeomProofs/Quadanglebisect.T[P]

12.1.11 Japanese Temple Problem

Given a circle and a chord through that
circle, select a point X on the chord.
There are two circles that are mutually
tangent to the chord at X and also tangent to the larger circle. Show that the
ratio of the radii of those two circles
is independent of the location of X.
What if X is on the chord but outside
the circle?

O2
r2
X
r1
O1

File: GeomProofs/Temple.T[P]

12.1.12 Von Aubels Theorem

D

N
O

A
M

B
P

Given a quadrilateral ABCD, if

squares are erected on the exterior
of each side, show that the line segments connecting the centers of the
squares erected on opposite sides have
the same length and are perpendicular.
File: GeomProofs/Vonaubel.T[P]

C
W

284

CHAPTER 12. GEOMETER PROOFS

12.2 Proof Exercises

This section contains a collection of problems, some easy, and some quite difficult.
Most of them are more difficult than problems found in standard high school geometry
textbooks. More difficult problems are preceded by the warning symbols: , or .
The solutions appear in the Geometer files corresponding to the code in front of each
file like E15 or C21. For example, to find the solution to problem E15, view the file
E15.T in the GeomProofs directory.

1. [E1]  If four straight lines emanate from a point, and if a line that intersects
all four lines is divided harmonically by the points of intersection, then any line
that intersects all four lines will also be divided harmonically by its points of
intersection.
2. [E2] If a line is tangent to a circle at point P and parallel to a chord AB of that
circle, then the point of tangency P will be the midpoint of the arc AB cut off
by the chord.
3. [E3]  Given two points A and B outside a circle K, show that if point C lies on
the circle, then the sum AC + BC is minimized when AC and BC make equal
angles with the tangent to K at C.
4. [E4] If circle K has center O, and circle K1 has diameter OR, where R is on K,
and let L be a line from R that cuts both K and K1 at P and P1 , respectively.
Show that RP/2 = RP1 .
5. [E5] If two circles (possibly of different sizes) cut each other at two points, show
that the line segment joining their points of intersection is bisected by the line
joining their two centers.
6. [E6] If K and K1 are two concentric circles with K1 lying inside K, and if a
line L is tangent to K1 at a point P , and L cuts K at points A and B, then
AP = BP .
7. [E7] Suppose that K1 and K2 are two circles that do not intersect, and neither
lies inside the other. The two common internal tangent lines and the two common
external tangent lines meet in four points. Show that these four points all lie on
a circle.
8. [E8] Let K1 and K2 be two circles which are tangent internally at a point P . Let
L be the common tangent line to the two circles at P . Let O be any point on L,
and let K be a circle centered at O that cuts both K1 and K2 . Show that the lines
connecting O to the points of intersection of K and K1 and K and K2 cut off
chords of K1 and K2 that are all of equal length.
9. [E9] Given a circle K centered at O, construct the two lines tangent to K at
the endpoints of some diameter. Let P be any point of K other than one of the
endpoints of that diameter. The line tangent to K at P intersects the other two
tangent lines at A and B. Show that OP OP = P A P B.
10. [E10] If two circles K1 and K2 having diameters d1 and d2 are externally tangent, and L is a line that is externally tangent to both K1 and K2 at points P1
and P2 , then P1 P2 P1 P2 = d1 d2 .

12.2. PROOF EXERCISES

285

11. [E11] Show that any side of a triangle is longer than the difference of the lengths
of the other two sides.
12. [E12] Show that in any right triangle, the line segment connecting the right angle
to the midpoint of the hypotenuse is equal to half the hypotenuse.
13. [E13] If P is any point in an equilateral triangle, and perpendiculars are dropped
from P to each of the three sides, intersecting the sides at points A, B, and C,
show that P A + P B + P C is equal to the altitude of the triangle.
14. [E14]  Let 4ABC be an isosceles triangle with ABC = ACB. Let D and
E be points on AB and AC such that DB = EB. Let P be the intersection of
lines BC and DE. Show that CEB = 2DP B.
15. [E15] Show that for any 4ABC, let P be the point that is the intersection of the
exterior angle bisectors at B and C. Show that P A bisects BAC.
16. [E16] Show that if the opposite angles of a quadrilateral are equal, then the
quadrilateral is a parallelogram.
17. [E17] Show that a line connecting opposite vertices of a parallelogram divides
the parallelogram into congruent triangles.
18. [E18] Show that if ABCD is a parallelogram, and P is any point interior to
ABCD, then A(4AP B) + A(4DCP ) = A(4BP C) + A(4DP A).
19. [E19] Let ABCD be any convex quadrilateral, and let P be a point interior to
ABCD. Show that AP + BP + CP + DP AC + BD.
20. [E20] If the sides of a regular pentagon are extended to form a five-pointed star,
show that the sum of the angles formed by the five tips of the star add to two
right angles. What if the pentagon is not regular?
21. [E21] Let 4ABC be a right triangle with the angle at A being the right angle.
Let P be any point on the segment AB. Show that CP 2 + AB 2 = BC 2 + AP 2 .
22. [E22] Let M be the centroid (the point where the three medians intersect) of
4ABC. Show that AM 2 + BM 2 + CM 2 = (AB 2 + BC 2 + CA2 )/3.
23. [E23] Let ABCD be a rectangle, and let P be any point in the interior of
ABCD. Show that P A2 + P C 2 = P B 2 + P D2 .
24. [E24] Given any 4ABC, let P be an arbitrary point inside the triangle, and
construct lines through P parallel to all three sides of 4ABC. If the line parallel
to AB intersects BC at point E and AC at point I, the line parallel to BC
intersects AB at J and AC at F , and if the line parallel to AC intersects AB at
G and BC at H, show that P E P F P G = P H P I P J. Is it necessary for
P to be inside 4ABC for this to hold?
25. [E25]  If 4ABC is a right triangle with right angle at A, and d is the diameter
of the inscribed circle of 4ABC, show that d = AB + AC BC.
26. [E26] Let K be a circle and P be a point outside K. The two tangent lines from
P are tangent to K at points A and B. Let C be any point on the arc from A to
B, and construct a tangent to K at C. The three tangent lines form a triangle.
Show that the perimeter of the triangle is P A + P B.
27. [E27]  Let K be the inscribed circle of any 4ABC, and let I be the center of
K. From B, drop a perpendicular to line AI, intersecting it at E. Show that E
lies on the line connecting the points of tangency of K with BC and AC.

286

CHAPTER 12. GEOMETER PROOFS

28. [E28] Let line L1 intersect a circle K at A and B, and let line L2 intersect K at
C and D. If L1 L2 , then show that AC 2 + BD2 = AD2 + BC 2 .
29. [E29] Let ABCD be a cyclic quadrilateral. Let G be the intersection of its
diagonals. Let GH AD with H on AD, and GI BC with I on BC. Show
that GH : GI = AD : BC.
30. [E30] Let K be a circle, and let A and A0 be inverse points with respect to K,
and similarly, B and B 0 are another pair of inverse points with respect to K.
Show that A, A0 , B, and B 0 either all lie on a straight line, or ABB 0 A0 is a
cyclic quadrilateral.
31. [E31] If O is the incenter of an arbitrary 4ABC, and X and Y are the points of
tangency of the incircle with edges AB and AC of 4ABC, show that O, X, Y ,
and A all lie on a circle.
32. [E32] Given an equilateral triangle, divide one side arbitrarily into two pieces
of length x and y. Construct two more equilateral triangles using the pieces of
lengths x and y as one side of each new triangle, and construct the new triangles
so that both are on the outside of the original triangle. Show that the triangle
connecting the centers of the three equilateral triangles is itself an equilateral
triangle.
33. [E33] If A and B are points at opposite ends of the diameter of a circle, and C
is any point on the circle, show that AC 2 + BC 2 is a constant.
34. [E34] Let AB be the diameter of a semicircle, and let C be a point on AB.
The perpendicular to AB through C intersects the semicircle at D and let E be
the midpoint of the semicircle. If r is the radius of the semicircle, show that
CD2 + CE 2 = 2r2 .
35. [E35]  Let AB be the diameter of a circle, and let CD be a chord of the circle
parallel to AB, where C is closest to A. Let X be any point between A and B.
Show that XC 2 + XD2 = XA2 + XB 2 .
36. [E36] Suppose that the diameter of a circle is divided into a large number of
pieces, and that circles are constructed with each of those pieces as its diameter.
Show that the sum of the circumferences of the smaller circles is equal to the
circumference of the original circle.
37. [E37] If three congruent circles are mutually tangent, what is the ratio of the area
of the triangle connecting their centers with the area of the triangle that connects
their points of tangency?
38. [E38]  Let K be a circle and let P be a point outside K. From P draw the two
tangents to K having points of tangency A and B. Let CD be any diameter of
K. If AC and BD intersect at a point Q, show that P Q CD.
39. [E39] If K1 and K2 are two circles that are externally tangent to each other, and
AB and CD are parallel diameters of K1 and K2 , show that AD and BC pass
through the point of tangency of K1 and K2 .
40. [E40]  Let D be the midpoint of AC in 4ABC, and let E be any point on
BC. The line DE intersects AB at G, and intersects the line through B that is
parallel to AC at point F . Show that H(GE, DF ). In other words, show that
the four points G and E divide D and F harmonically.

12.2. PROOF EXERCISES

287

41. [E41] Given an angle, construct its bisector and a line perpendicular to the bisector passing through the vertex of the angle. Let L be any line intersecting all four
of the lines in the figure. Show that this L intersects the four lines in a harmonic
set.
42. [E42] Given two circles of radius 1 tangent to each other, and each tangent to
a line that does not intersect the circles at their point of tangency. What is the
radius of a circle that is tangent to both circles and to the line?
43. [E43] Given two circles that do not intersect and such that one does not lie inside the other. Show that the four points where the common external tangents
touch the circles lie in a circle K1 . Also show that the four points where the
common internal tangents touch the circles lie in another circle K2 . Show that
the four points of intersection of the common external tangents and the common
internal tangents also lie on a circle K3 . Finally, show that K1 , K2 , and K3 are
concentric.
44. [E44]  Given two circles that do not intersect and do not lie one inside the
other, if the common internal and external tangents are constructed, show that
the length of an internal tangent is equal to the distance between the points on
the external tangent where it is intersected by the two internal tangents.
45. [E45]  Given an arbitrary 4ABC. Let D lie on BC such that AD bisects
BAC. Show that AD 2 = AB BC BD CD.
46. [E46] If 4ABC is a right triangle with the right angle at A, and a perpendicular
to BC is dropped from A that meets BC at D, show that AB 2 : AC 2 = BD :
BC.
47. [E47] Given a circle K and a point A outside of K, let AX and AY be the two
tangents from A to K, where X and Y lie on K. Let P be any point on K, and
extend line AP to Q on K. Show that XP/XQ = Y P/Y Q.
48. [E48] Given any 4ABC, show that it is possible to construct a rectangle with
one vertex at A, and the other three vertices on the circles whose diameters are
AB, AC, and BC.
49. [E49]  Suppose hexagon ABCDEF is inscribed in a circle, and the diagonals
AD, BE, and CF meet at a point. Show that AB CD EF = BC DE F A.
50. [E50]  Suppose a circle intersects all three sides of 4ABC twice. It intersects
AB at C 0 and C 00 , it intersects BC at A0 and A00 , and it intersects CA at B 0 and
at B 00 . Suppose further that the lines AA0 , BB 0 , and CC 0 meet at a point. Then
the lines AA00 , BB 00 , and CC 00 also meet at a (different) point.
51. [E51]  Given 4ABC and line L, let L intersect the sides of AB, BC, and
CA at C 0 , A0 , and B 0 , respectively (some or all of the intersections may be
outside the triangles edges). Show that the midpoints of the segments AA 0 ,
BB 0 , and CC 0 all lie on a straight line. (This line is called the Gauss line.)
52. [E52]  Given a right triangle 4ABC, where B = 90 , construct a circle
internally tangent to the legs AB and BC, and to the arc AC. Show that the
diameter of this circle is twice the diameter of the inscribed circle of 4ABC.
53. [E53] Given 4ABC, let A0 , B 0 , and C 0 be the feet of the altitudes opposite
A, B, and C, respectively. If perpendiculars are dropped from A 0 to AB, AC,
BB 0 , and CC 0 meeting them at K, N , L, and M , then K, L, M , and N lie in a
straight line.

288

CHAPTER 12. GEOMETER PROOFS

AB CD EF GH IJ

= 1.
BC DE F G HI JA

54. [E54]  Given two squares ABCD and AEF G with centers P and Q that share
a common point A, let J and K be the midpoints of the segments DG and BE,
respectively. Show that P JQK is a square. This is called the Finsler-Hadwiger
theorem.
55. [E55] Pick four points A, B, C, and D in clockwise order on a circle. Let W ,
X, Y , and Z be the midpoints of the arcs AB, BC, CD, and DA, respectively.
Show that W Y XZ.
56. [E56]  Let M be the midpoint of the longer arc CD of a circle K. Pick a point
X at random on K. Construct XC, XD, and drop a perpendicular from M to
XD, meeting it at Y . Show that XC + XY = Y D. This is called Archimedes
midpoint theorem.
57. [E57]  In any 4ABC, let M be the midpoint of AC and let M N be parallel to
the angle bisector of ABC, where N is the other intersection with the triangle
(on either AB or BC, depending on the shape of 4ABC). Show that M N
divides the perimeter into two equal areas. (Hint: See problem 56.)
58. [E58] If an altitude of a triangle is extended to intersect the circumcircle, then
the length of the segment from the orthocenter to the base of the triangle is equal
to the length of the segment from the base to the circumcircle.
59. [E59] In 4ABC if the excircle opposite A touches the side BC at D, then
AD bisects the perimeter of 4ABC.
60. [E60] Given two circles and a line that are all tangent at a common point X
with both circles on the same side of the line, pick any other point on that line
and draw the other tangent to the inner circle. That line intersects the outer
circle twice, at A and B, and is tangent to the inner circle at P . Show that
AXP = P XB.
61. [E61] Show that if BAC = 30 in 4ABC, then BC is equal to the length of
the radius of the circumscribed circle of 4ABC.
62. [E62] If circle K has its center on another circle, then the common chord of
the two circles is equal to the length of the chord of K passing through the
intersection of the two circles and tangent to the circle on which its center lies.
63. [E63] Let A, B, and C be three non-collinear centers of circles, none of which
intersect, and none of which lie inside another. Let A0 , B 0 , and C 0 be the intersections of the common internal tangents to the circles centered at B and C, at
C and A, and at A and B, respectively. Show that AA0 , BB 0 , and CC 0 meet at
a point.
64. [E64]  Given 4ABC, construct two right triangles 4BAD and 4BCE on
the outside of 4ABC such that ABD = CBE. Let F be the foot of the
altitude from B to AC. Show that the lines AE, CD, and BF are concurrent.
65. [E65] Draw a pentagram (a five-pointed star) by choosing five points A, C, E,
G, I and forming the segments AE, EI, IC, CG, and GA. Let AE and IC
intersect at B, CG and AE at D, EI and CG at F , GA and EI at H, and IC
and GA at J. Show that

12.2. PROOF EXERCISES

289

66. [E66] Construct a square on the hypotenuse of a right triangle and draw a line
from the vertex opposite the hypotenuse to the center of the square. Show that
this line splits the right angle into two 45 angles. Does it matter on which side
of the hypotenuse the square is constructed?
67. [E67] Let 4ABC be a right triangle with ABC = 90 . Let AB = 3 BC.
If M is the midpoint of AC, and D and E trisect AB, show that 4DM E is an
isosceles right triangle.
68. [E68] A circle with diameter P P 0 is tangent to a line L at P 0 . Let XY be a chord
of the circle parallel to P P 0 . If Q is the midpoint of the arc XY , and if lines
P X, P Q and P Y are extended to intersect L at X 0 , Q0 , and Y 0 , respectively,
show that H(P 0 Q0 , X 0 Y 0 ).
69. [E69]  If a quadrilateral is circumscribed around a circle and if the lines connecting the opposite points of tangency are perpendicular, then the quadrilateral
can be inscribed in a larger circle.
70. [E70] If the altitudes of 4ABC are ha , hb , and hc , and the radius of the inscribed circle is r, show that 1/ha + 1/hb + 1/hc = 1/r.
71. [E71] If ABCD is a trapezoid with AB k CD and squares are erected on the
exterior of sides AB and CD, the line connecting the midpoints of the squares
passes through the intersection of lines AC and BD.
72. [E72] Let I be the incenter of 4ABC. If O is the circumcenter of 4BIC,
show that A, I, and O lie in a straight line.
73. [E73] In a semicircle with center O inscribe a circle K tangent at O. Inscribe a
second circle K1 in the semicircle tangent to the semicircle, its diameter, and to
K. What is the ratio of the radius of K1 to the radius of the semicircle?
74. [E74] Three circles of radius 1 are tangent to each other and their diameters are
all part of the same line. A line is drawn from the left end of the diameter of the
leftmost circle which is tangent to the rightmost circle. This tangent line cuts a
chord from the middle circle. Find the length of that chord.
75. [E75] Given two circles of radius R and r tangent to each other, and both tangent
to a line that does not pass through their point of common tangency, construct
a circle tangent to both circles and the line. Find the radius of the new circle.
Suppose that you begin with two circles of radius 1, inscribe the circle between
them and their tangent line, then inscribe circles between both pairs of the three
circles, then between all pairs of the set of 5 circles, and so on. What are the
radii of the circles so generated? In fact, if two adjacent circles have radii 1/m 2
and 1/n2 , where m and n are integers, whats the radius of the new circle? (See
Exercise 42.)
76. [E76] Given two triangles 4ABC and 4DEF inscribed in a circle such that
the vertices alternate: A, D, B, E, C, F , show that an ellipse can be inscribed in
the interior hexagon. What if the two triangles are inscribed in an ellipse instead
of in a circle?
77. [E77] In 4ABC, let A0 , B 0 , and C 0 lie on lines BC, CA, and AB, respectively
such that AA0 , BB 0 , and CC 0 meet at a point. Show that

 
 

sin BAA0
sin CBB 0
sin ACC 0

=1
sin CAA0
sin ABB 0
sin BCC 0
.

290

CHAPTER 12. GEOMETER PROOFS

78. [E78] In 4ABC, let A0 , B 0 , and C 0 lie on lines BC, CA, and AB, respectively
such that AA0 , BB 0 , and CC 0 meet at a point P . If the lines AA0 , BB 0 , and CC 0
are reflected across the corresponding interior angle bisectors, those reflected
lines will also meet at another point Q. Points P and Q are called isogonal
conjugates. (See Exercise 77.)
79. [E79] Show that in any 4ABC, the orthocenter and the circumcenter are isogonal conjugates. See Exercise 78 for the definition of isogonal conjugate.
80. [E80] Show that the diameter of the nine-point circle of a triangle is half the
diameter of the circumcircle of that triangle.
81. [E81] Show that if a ray is drawn from the orthocenter H of a triangle that
intersects the nine-point circle at P and the circumcircle at Q, that HP = P Q.
82. [E82] Let AA0 , BB 0 , and CC 0 be the altitudes of 4ABC, where A0 , B 0 , and
C 0 lie on BC, CA, and AB, respectively. Show that AA0 , BB 0 , and CC 0 are
the angle bisectors of 4A0 B 0 C 0 .
83. [E83] Let P , A, B, and C lie on a circle. Construct the three circles with diameters P A, P B, and P C. In pairs, those three circles meet in three points other
than point P where they all meet. Show that those three intersection points lie
on a straight line.
84. [E84] If three circles of equal radius centered at A, B, and C all pass through
the same point O, and no pair are tangent, let A0 , B 0 , and C 0 be the pairwise
intersections of the circles centered at B and C, at C and A, and at A and B,
respectively, show that AA0 , BB 0 , and CC 0 meet at a point.
85. [E85] Let A, B, C, D and E be 5 concyclic points let A0 , B 0 , C 0 , D0 and E 0 be
the midpoints of CD, DE, EA, AB and BC, respectively, and let V , W , X,
Y and Z be the midpoints of B 0 E 0 , A0 C 0 , B 0 D0 , C 0 E 0 and E 0 A0 , respectively.
Show that V , W , X, Y and Z are concyclic.
86. [E86] Let A, B and C lie on a line with C between A and B such that AC =
3CB. Two circles are drawn with diameters AC and CB. The common exterior
tangents of those two circles meet at D. Show that 2BD = BC.
87. [E87] Let 4ABC contain only acute angles. Construct circles having diameters
AB, BC and CA. From each vertex of the triangle, draw the two tangent lines
to the circle whose diameter is the opposite side. Show that the six points of
tangency all lie on a circle.
88. [E88] Show that if the angle bisectors of a triangle are equal, then the triangle
is isosceles. This is the Steiner-Lehmus theorem.
89. [E89] If two circles intersect at points A and B and a line that passes through
the segment AB intersects the circles at the points W , X, Y and Z in that order,
show that ZAY = W BX.
90. [E90] If the quadrilateral ABCD has perpendicular diagonals and if W , X, Y
and Z are the midpoints of the sides, then if perpendiculars are dropped from W ,
X, Y and Z to the points W 0 , X 0 , Y 0 and Z 0 on the opposite side, then all eight
points W , X, Y , Z, W 0 , X 0 , Y 0 and Z 0 lie on the same circle. This is called the
eight-point circle theorem.
91. [E91] If a circle is drawn whose diameter is the segment connecting the incenter
with one of the excenters, then that circle passes through two vertices of the
triangle. This is called the incenter-excenter theorem.

12.2. PROOF EXERCISES

291

92. [E92] Let A be a point outside circle O and choose a point P such that the
tangent P T to the circle has the same length as P A. If C is any point on O, let
B and D be the intersections of P C and CA with O, respectively. Let E be the
intersection of AB with O. Show that DE k AP .
93. [E93] Show that if the cevians AD, BE and CF of 4ABC are concurrent, then
the other intersections of the triangle with the circle passing through D, E and
F are the endpoints of another set of concurrent cevians.
94. [E94] Show that if two chords of a circle are perpendicular, then the sum of the
squares of the chord segments is equal to the square of the diameter of the circle.
95. [E95] If three circles of equal radius R meet at a point, then the pairwise intersections of those circles lie on a circle that also has radius R. This is known as
Johnsons theorem.
96. [E96] Given two lines AB and AC, choose D and E to lie on lines AB and
AC, respectively. Let F be the intersection of CD and BE. Show that the four
circumcircles of 4ABE, 4ADC, 4BDF and 4CEF are concurrent. (Hint:
consider the Simson line.)
97. [E97] Show that if 4ABC is a right triangle with C = 90 , then the altitude
through C is equal to the median through C reflected across the angle bisector
through C. Notice that this proves that the three symmedians of a right triangle
meet on the altitude through the right angle, since we know that the three symmedians meet at a point and this exercise shows that one of the symmedians is
the altitude through C.

292

CHAPTER 12. GEOMETER PROOFS

Appendix

Mathematics Review
This book is about solving geometric problems, so this chapter reviews mainly geometric topics, but not entirely. Vectors, complex variables, and analytic geometry, for
example, can sometimes provide quicker and better solutions for problems that at first
appear to be completely geometric.
If you feel even a little rusty with respect to high school mathematics, it may be
worthwhile to skim through this appendix to get a feeling for what is important.
Since this appendix is a review, in most cases proofs of the results are not included.
However, many of the more important theorems do exist as proofs in Geometer files.
These may be useful for teachers of geometry to present the proofs to a class in a
more interactive way. When such a proof exists for a particular result, it can be found
in the directory Proofs on the CD, and a reference to it will be included in the text
accompanying the statement of the result in this chapter.
No attempt has been made to present the material in this chapter in a logical
order such that each topic is defined and discussed before it is used. The ordering here
293

APPENDIX A. MATHEMATICS REVIEW

294

is designed to show the reader the relationships among the various parts of geometry.
Almost any high school geometery text can be consulted for a logically consistent
order.

A.1 Types of Geometric Problems

Here is a list of the typical sorts of problems that are posed in geometry. It is difficult
to divide geometry problems cleanly into sets, but here are some general categories.
Finding the measure of something. This is why geometry was inventedgiven
some measurements (lengths, angles, areas), calculate some other measurements.
The most useful tools to solve these sorts of problems are formulas like the
pythagorean theorem, the law of sines and the law of cosines, Stewarts theorem, the formulas for areas of objects, et cetera, but there are other techniques
available.
Geometric constructions. The most famous of these problems require constructions of geometric objects given certain initial objects and a straightedge and
compass1 . Sometimes construction with other sets of allowable tools is studied,
but the most famous set consists of a straightedge and compass. Construction
problems are somewhat artificial, but the exercise of solving them can be very
beneficial to ones geometric intuition.
Finding and proving relationships. Quite often these problems simply require
that you prove some relationship, but problems that require you to study a situation to find the relationship first and then to prove it are more instructive.
The word relationship can mean many things. Here are a few examples. Are
certain lengths or areas or ratios or angles equal? Is one always greater than the
other? Are two figures congruent or similar? Do three lines or circles meet at a
point, or do four points lie on a circle? Are two lines parallel or perpendicular?
Is a line tangent to a circle? Are two circles concentric or orthogonal?
Finding a locus of points.

Everyone knows that a circle is the set of points at a fixed distance from a given
point, but what is the set of points equidistant from a point and a line? Given a
circle and a point, what is the locus of points at the intersection of a line tangent
to a circle and the line perpendicular to it passing through the point? Such easilystated problems can lead to the study of fantastic curves.

A.2 Congruence
1 Mathematicians assume that the tools are ideal in the sense that they can be used to draw infinitely
accurate lines and circles.

A.2. CONGRUENCE
Congruence is technically an undefined term. In
theory all we know about congruence is from
its properties as stated in the postulates, but it
is easy to understand intuitively. Two geometric figures are congruent if they are exactly the
same shape and size. In other words all measurements are the same, be they lengths or angles. If five lines meet at a point in one figure,
the corresponding five lines meet at the corresponding point at the same angles in a congruent figure, et cetera.
Imagine two geometric figures drawn on
transparent plastic. They are congruent if one of
the pieces of plastic can be placed on top of the
other in such a way that the two figures match
exactly. When doing this matching, it is OK to
turn one of the sheets upside-down.

295

C
C

B
A

C
A

Figure A.1: Congruent Triangles

Review/Congruent.T [M]

 If a sheet can be flipped to make a match, the two figures are said to be mirror
images of each other. Notice that if a figure has certain symmetry properties, it can be
a mirror image of itself. The concept of congruence is not limited to two-dimensional
figures; the same idea makes sense in three (or more) dimensions, but it is harder to
imagine drawings on three dimensional plastic sheets and especially in making them
coincide. But what is especially hard to imagine is turning these three dimensional
sheets over to obtain the mirror image so that, for example, your left hand and right
hand would be approximately congruent. The flipping of this three-dimensional sheet
of plastic would require movement through a fourth dimension.
is used to indicate congruence. Figure A.1 shows four congruent
The symbol =
triangles. When we list the vertices in the equation that shows the congruence, we list
them in the same order. Thus 4ABC
= 4A0 B 0 C 0 , but it is not true that 4ABC
=
0 0 0
4B A C . This second formula would indicate that A in the first triangle matches B 0
in the second, B matches A0 , and so on.
If the geometric figures have certain amounts of symmetry, there may be multiple
ways to set up a congruence. Every figure, of course, is congruent to itself, but for
something like an equilateral triangle, there may be many self-congruences. If 4ABC
is equilateral, then we have:
4ABC
= 4BCA
= 4CAB
= 4ACB
= 4CBA
= 4BAC.
A regular hexagon would be congruent to itself in 12 ways.
 Notice in figure A.1 that although they are congruent, 4ABC and 4A 000 B 000 C 000
are mirror images of each otherit is impossible to move one of them in the plane
until it lies exactly on top of the other; one of them must be flipped over before
they can be made to match. Notice also that when a triangle is flipped, the clockwise/counterclockwise ordering of the points reverses. In 4ABC the points A, B, and
C appear in a clockwise orientation; in 4A000 B 000 C 000 the corresponding points A000 ,

296

APPENDIX A. MATHEMATICS REVIEW

B 000 and C 000 appear in a counterclockwise order. Any mirror image of an object will
convert clockwise to counterclockwise orientations and vice-versa.
Although high-school geometry books concentrate on the congruence of triangles,
any sorts of geometric figures can be compared to see if they are congruent. Two line
segments are congruent if they are the same length. Two circles are congruent if they
have the same radius, two angles are congruent if they have the same measure, et cetera.
Congruence is called an equivalence relation because it has the following three
properties:
1. The relation is reflexive; every figure is congruent to itself: A
= A.
2. The relation is symmetric: if A
= B then B
= A.
3. The relation is transitive: if A
= B and B
= C then A
= C.

A.2.1 Similarity
Similarity is like congruence, but possibly using a magnifying glass. Two geometric
figures are similar if they have the same shape, but not necessarily the same size. In
other words, two figures are similar if it is possible to make them congruent by magnifying one of them.
The symbol is used to indicate similarity. If 4ABC is similar to 4DEF , we write
4ABC 4DEF . In the same way that the
vertices of congruent objects are listed in a corresponding order, so are the vertices of similar objects. In the example above, this means
that if we magnified 4DEF to a new triangle
4D0 E 0 F 0 , then 4D0 E 0 F 0
= 4ABC, where
D0 corresponds to A, E 0 to B, and F 0 to C.
Similarity, like congruence, is an equivalence relation. Remember that congruent objects are also similarthe magnification is by
a factor of 1.

D
A

s
C

v
B

Similarity is used more often than congru- Figure A.2: Proof of Chord Theoence. If it is known that two figures are similar, rem
then all the ratios of the corresponding lengths
Review/Crossproof.T [P]
are the same. In other words, if one drawing is
twice as big as another (in length), then every
line segment in the first drawing is exactly twice as long as the corresponding segment
in the other. In almost any geometric problem where there is a product of lengths or a
ratio of lengths involved, the solution will probably follow from similarity arguments.
As an example of this technique, consider figure A.2. Suppose we are given a
circle with two lines passing through a point P inside it. If A, B, C and D are the

A.2. CONGRUENCE

297

intersections of the lines with the circle as in the figure, then AP P B = CP P D (or
equivalently, st = uv).
If we are going to show that st = uv using similarity arguments, we had better form
some triangles and there are only a few possibilities. Here we construct the segments
AC and DB although it also works to connect AD and BC. Similarity lets us deal with
ratios, so if st = uv that is equivalent to s : v = u : t (or, if you prefer, s/v = u/t).
This will be true if 4BDP 4CAP .


Two triangles are similar if they have two equal angles. BP D = CP A because
they are vertical angles. Also, CAP = BDP since they both are inscribed in the
same arc CB. Therefore 4BDP 4CAP , so s/v = u/t, so st = uv.
The entire field of trigonometry is based on similarity. See Section A.10.1

A.2.2 Measure
The word geometry is derived from the Greek  which comes from ,
the combining form of (earth) plus  (measure). The original application
of geometry was to measure the earthto find areas of fields, distances between cities,
and so on.
In spite of this, in pure Euclidean geometry, there is no such thing as an absolute
measure of anything. Two line segments can be compared with the result that
they are
the same length, or that one is twice as long as the other, or that one is 2 times as
long as the other, but the process is always one of comparison. There is no particular
segment that is one inch or one meter long. If a diagram for a theorem in Euclidean
geometry is put in a copy machine with magnification on, the resulting diagram will
illustrate the theorem equally well.
Geometric figures are scale drawings where the scale is never stated. If a diagram
contains line segment A that is twice as long as line segment B, then if A is two inches,
B is one inch, or if A is two centimeters, B is one centimeter, or if A is two light-years,
B is one light-year, et cetera. All the measurements in a diagram are relative.
Of course it is possible to talk about ratios of areas (and of volumes, in solid geometry) but the types must match in comparisons. Areas can be compared with areas,
volumes with volumes, et cetera. Just as in physics you cannot compare a time with a
distance you cannot compare a geometric area with a volume. You can compare ratios
like area : length = area : length, but notice that the types on both sides are the
samean area to length ratio in this case.
In physics, a velocity can be measured in feet per second, miles per hour, or furlongs per fortnight. All these are different units, but they all have something in common: each is a length divided by a time, and all can be interconverted. But a distance
cannot be converted to a velocityone has units of length and the other has units of
length divided by time.
Similarly, in geometry, lengths cannot be compared to areas or volumes, since areas
have units of length squared and volumes of length cubed. A ratio of two lengths
has units of length divided by length, so it is dimensionless, and cannot be compared to

APPENDIX A. MATHEMATICS REVIEW

298

t
v

v
s

Figure A.3: The Chord Theorem II

Review/Cross.D [D]

lengths or areas. But a ratio of an area to a length is just a lengthits units are length
squared divided by length, or just length.
Statements about ratios of lengths can be converted to equivalent statements about
areas by cross-multiplying. In the previous section we proved the chord theorem which
states that st = uv (see figure A.3). The same theorem can equally well be stated as an
equality of ratios: s : v = u : t (which is equivalent to s/v = u/t, unless v = t = 0).
Multiply this equation on both sides by v and t to obtain the area equality s t = u v.
A Geometer proof of the chord theorem can be found in Proofs/CrossProd.T and
Proofs/Crossproof.T.
Of course the product of two lengths can be interpreted as an areathe area of a
rectangle whose sides are those two lengths. Saying that s t = u v is equivalent to
stating that two rectangles, one with sides s and t and the other with sides u and v, have
the same area. Those two rectangles are shown in the right part of figure A.3.
This idea of converting a ratio equality to an area equality or vice-versa can be
extremely powerful. Ratios arise very naturally from similarity (see Section A.2.1), so
if you need to prove something about the product of two lengths and you do not have
any idea how to proceed, convert those products into ratios and see if you can show
that they are parts of similar geometric figures2 .
A detailed examination of the concept of area is beyond the scope of this book.
Even if we say that the area of a circle having radius r is r 2 we have moved beyond
simple geometry to calculuswhat does it really mean to talk about the area of a
geometric figure with curved edges when area was originally defined only for shapes
whose boundaries are straight lines?
Like ratios, angles are dimensionless. The 30 in a 30 angle is effectively part
of a ratio: it is 30 /360 = 1/12 of a 360 circle. In radians, the same angle is
still 1/12 of a circle, but this time the measure of /6 is in relation to a circle whose
circumference is 2. The unit of measure of an angle is simply a comparison to the
somewhat arbitrary measure that is assigned to a full circle. The pros and cons of
various angle measures are discussed in detail in Section A.5.1.
2 Sometimes it is convenient to consider positive and negative ratios (this is sometimes called a directed
ratiosee Section 5.2.1).

A.3. POINTS

299

A.3 Points
The term point itself is technically undefineda point is simply an object that happens
to satisfy the properties specified by the postulates. It is useful to think of a point as an
infinitely small dot or position on the plane (or in space, in the case of solid geometry).
Sometimes it is useful to think of other geometric figureslines, circles, parabolas,
et cetera, as collections (or sets) of points. In fact, a definition of a circle might be: A
circle is the set of all points that are equidistant from a given point. Or equivalently:
A circle is the locus of points equidistant from a given point.
A point has no sizezero length and zero area. (In Elements ([Euclid, 1956]),
Euclid tried to define the term as follows: A point is that which has no part.)
Here are some definitions of special types of points:
Midpoint. A point M is the midpoint of two other points A and B if it lies on the
line AB and AM = M B.
Collinear Points. A set of points is collinear if all of them lie on the same line.
Concyclic Points. A set of points is concyclic if they all lie on the same circle.
Point of Intersection. When two lines, or a line and a circle, or two circles (or various
other curves, like conics sections) have a point in common, that common point
is called the point of intersection. Depending on the figures, there may be
more than one point of intersection. Quite often the terms intersection of two
lines, intersection of two circles, et cetera, is used in place of the point of
intersection of two lines, the point of intersection of two circles, et cetera.
Point of Tangency. When a line is tangent to a curve (in other words, when the line
and the curve locally share a single point and have the same slope), that point
in common between the line and curve is called the point of tangency. See the
definition of tangency in the next section.
There are very few theorems that deal exclusively with points; almost all deal with
points in relation to other geometric figures, and those theorems are covered in the
sections that follow.

A.4 Lines
Like point, the term line is technically undefined, but most people have a good idea
what it signifies. In Euclidean geometry, line will refer to a straight line that extends
forever in both directions. It is infinitely thin, in the sense that if you pass across it,
its width is only a single point. A line is composed of an infinite number of points.
If A and B are two different points, the line AB refers to the particular line that
passes through those two points (the postulates guarantee that there is only one such

APPENDIX A. MATHEMATICS REVIEW

300

line). In some books the symbol AB is written to mean the line AB, where the arrow
above emphasizes that the line continues forever in both directions.
Given two points, the line segment connecting them is the shortest path between
them.
A common misconception is that one can somehow talk about points on that line
that are next to each other. Between any two different points on the line there are
other points (in fact an infinite number of other points).
Here are some definitions of special lines, and of portions of lines:

F
A

E
B

Figure A.4: Lines

Review/Lines.D [D]

Segment. A segment is a piece of a line, consisting of two points on that line (the
endpoints of the segment) and all the points that lie between those endpoints on
the line. Sometimes the segment AB is written as follows: AB. A segment has
a length, but its area is zero.
Ray. Intuitively, a ray is half of a linepick a point on a line and all the points on
one side of it and you have a ray. This can be made precise as follows: Given

a segment AB of a line L, the ray AB consists of all the points of the segment
AB, together with all the points C on L such that B lies between A and C. The

ray AB is said to emanate from point A.

Parallel Lines. Two lines are parallel if they lie in the same plane and have no points
in common. (In plane geometry there is only one plane, so we merely need to
say the lines have no points in common.) If L1 and L2 are two parallel lines,
we write L1 k L2 . In figure A.4, lines AB and CD are parallel. In Euclidean
geometry, parallel lines remain the same distance apart.
Transversal Line. Given a pair of parallel lines, a line that cuts through both of them
is said to be a transversal. In figure A.4, lines AC and EF are transversals to
the pair of parallel lines AB and CD.
Tangent Line. A line is tangent to a circle if it has exactly one point in common
with the circle. In other words, it just touches the edge of the circle. All the
points of the circle except for the point of tangency lie on one side of the line.
The general concept of tangency of a line and a curve requires calculus to define

A.4. LINES

301

Figure A.5: Tangent Line

Review/Tangent.D [D]

what is meant by the slope of a curve at a point. A line is tangent to a curve if it

passes through a point on the curve and the slope of the line is the same as the
slope of the curve at that point. Note that a line can be tangent at one point and
then pass through the curve at another or that the tangent line can pass through
the curve at the point of tangency. Figure A.5 illustrates a line tangent to a curve
at the point T that passes through the curve at another point.
Two curves (including circles, of course) are tangent at a point if they both are
tangent to the same line at that point.
Perpendicular Lines. Two lines are perpendicular if they form a right angle (a 90
angle). If L1 and L2 are perpendicular, we write L1 L2 . In figure A.4,
EF AB. (Of course EF CD as well.)
Perpendicular Bisector. A line that is perpendicular to a segment and divides the segment into two congruent segments is called the perpendicular bisector of the
segment. In figure A.4, point E is the midpoint of AB, so AE = EB. Since
line EF is perpendicular to line AB, then EF is the perpendicular bisector of
the segment AB.
Here are some key theorems about lines in plane Euclidean geometry. All can be
proved from the postulates but many of them are simply assumed in a standard high
school geometry course.
Two distinct lines can intersect in at most one point.
Every line contains an infinite number of points.
Given any four distinct points on a line, it is possible to label them A, B, C, and
D such that B lies between A and C and between A and D, and also that C lies
between A and D and between B and D.
Every line divides the plane into two regions such that if two points not on the
line are in the same region, then the segment connecting them does not intersect the line, and if the two points are not in the same region, then the segment
connecting them does have a point in common with the line.

302

APPENDIX A. MATHEMATICS REVIEW

A.5 Angles
It is a bit tricky to define the term angle.
See Section A.5.1 for a more thorough discussion. Here is
a usable
definition: Two
rays in the plane OA and OB both emanating from the same point O, and not lying
in the same line, form an angle (AOB or
BOA). The angle divides the rest of the
plane into two regions. A point P of the
plane is on the interior of AOB if a line
exists through P that intersects both rays
such that P lies between the two points of
intersection. If no such line exists then P
is exterior to the angle.

A
B
I

C
N

O
M

Figure A.6: Angles

Review/Angles.D [D]

Notice that this definition requires angles to be larger than 0 and smaller than 180. With this definition, straight angles
(angles measuring 180 ) and reflex angles (angles larger that 180 ) do not exist. The
discussion and definitions below sometimes ignore these problems.
Supplementary Angles. Two angles that add to a straight angle (to 180 ), are called
supplementary angles. They need not share a vertex. In figure A.6, M OJ
and JOL are supplementary.
Right Angle. An angle is a right angle if it is congruent to its supplementary angle.
Intuitively, right angles are made by perpendicular lines, or lines that meet at
90 . In figure A.6, GHI is a right angle.
Complementary Angles. Two angles are called complementary angles if they add
together to make a right angle. They need not share a common vertex. In figure A.6, GHN and N HI are complementary.
Vertical Angles. When two lines cross, four angles are formed, and the opposite pairs
are known as vertical angles. In figure A.6, LOK and M OJ are vertical
angles, as are LOJ and M OK.
Acute Angle. An acute angle is an angle that is less than a right angle. Without
using the concept of the measure of an angle, we can simply say that an acute
angle is an angle that can be obtained inside a right angle. In figure A.6, ABC,
GHN , N HI, LOK, and JOM are acute angles, among others .
Obtuse Angle. An obtuse angle is an angle that can contain a right angle. In other
words, an angle that is larger than 90 (and less than 180) is obtuse. In figure A.6, DEF , JOL, and KOM are examples of obtuse angles.

A.5. ANGLES

303

Angle Bisector. A ray emanating from the same point as the two rays defining an
angle that divides the angle into two congruent angles is called the angle bisec

tor. In figure A.6, the ray OP is the angle bisector of KOL. In other words,
KOP = P OL.
Straight Angle. A straight angle is not a proper angle as defined above, but you can
think of it as a 180 angle. In figure A.6, M OL is a straight angle.
Reflex Angle. A reflex angle is not a proper angle as defined above, but you can think
of it as the outside of the angle in question. In other words, a 30 angle will
have a reflex angle of 330 . In figure A.6, DEF and its reflex angle are shown
with one ring and two rings. They would both be written the same way: DEF ,
so there is clearly some deficiency in the naming convention that can be avoided
with directed angles (see Section A.5.2).

A.5.1 More on Angles

The previous definition of angle restricts the concept of angle to those having measures
between (but not including) 0 and 180 . For much of geometry, this is sufficient, but
a much more general definition of angle is often useful. It is sometimes nice to allow
angles of any sizepositive, negative, or zero.
A common approach is to define the measure of an angle as a sort of total turning
relative to a fixed ray. In the Euclidean plane with a standard Cartesian coordinate
system, that fixed ray is usually the positive x-axis, and positive angles are measured in
a counterclockwise direction from that. Negative angles, of course, represent clockwise
turning. Angles can thus have any real measure, positive, negative, or zero, and can be
added or subtracted without worryanother valid angle is certain to result.
If angles can have any positive or negative measure, one has to decide what is meant
by congruence. Is a 10 angle equal to a 370 angle? There is nothing particularly
tricky going on, but it is good to keep in mind that alternative definitions have both
pros and cons.
In this text, we will measure angles in degrees instead of radians. This is in line
with what appears in most geometry textbooks, and it is easy to convert back and forth.
To convert between degrees and radians, radians is equal to 180 degrees.

A.5.2 Directed Angles

If we wish to refer to reflex angles there is a labeling problem since an angle and its
reflex angle share the same vertices. For example, in figure A.6, the points D, E, and
F in reality determine two angles, one of which is marked with a single ring and the
other with two rings. If one has measure , the other has measure 360 , but both
would be written DEF .
One way around this is the idea of directed anglesassume that DEF im

plies that the ray ED is the first one, and the ray EF is the second. The angle goes

APPENDIX A. MATHEMATICS REVIEW

304

counterclockwise from the first ray to the second so in figure A.6, F ED is the one
that is less than 180 and DEF would be its reflex anglean angle larger than 180.
We seldom encounter situations where this distinction is necessary, but the distinction
can be made. For a more complete discussion, see [Johnson, 1917].

A.6 Triangles
If A, B, and C are any three non-collinear points, then the triangle 4ABC consists of
the three segments AB, BC, and CA. The points A, B, and C are called the vertices
of the triangle and the angles ABC, BCA, and CAB are its angles. Sometimes
it is useful to think of a triangle as the union of the three lines of which the segments
are parts. A triangle is just a special case of a polygon, so some general polygon
definitions, like perimeter and area, for example, will apply to triangles as well.
Equilateral Triangle. If the three segments composing a triangle are congruent, the
triangle is equilateral.
Isosceles Triangle. If two of the three segments composing a triangle are congruent,
the triangle is an isosceles triangle. The third side does not necessarily have to
be differentan equilateral triangle is a special case of an isosceles triangle.
Scalene Triangle. A scalene triangle is a triangle none of whose segments are congruent. Equivalently, no pair of its three angles are congruent.
Right Triangle. If two of the segments composing a triangle are perpendicular, it is
called right triangle. The two sides forming the right angle are called legs
and the third side is called the hypotenuse. Although most right triangles are
technically scalene triangles, they are hardly ever referred to in that way.
Acute Triangle. If the three angles of a triangle are all acute (less than 90 ), it is called
an acute triangle.
Obtuse Triangle. If one of the angles of a triangle is obtuse (greater than 90 ), it is
called an obtuse triangle.
Median. A segment connecting a vertex of a triangle to the midpoint of the opposite side is called a median of the triangle. Every triangle has three medians
and they meet at a point called the centroid, center of gravity, or center of
mass. The median of 4ABC in figure A.7 is G. A Geometer proof appears in
Proofs/Medians.T

Centroid. The intersection of the three medians. The centroid is 2/3 of the distance
from any vertex to the midpoint of the opposite side.
Inscribed Circle. A circle inscribed inside a triangle and tangent to all three sides. It
is also called the incircle.

A.6. TRIANGLES
B

305

C
G

H
A

Figure A.7: Medians, Angle Bisectors, and Altitudes

Review/Tricenters.T [M]

Figure A.8: Inscribed and Circumscribed Circles

Review/Trianglecircs.T [M]

Incenter. The three angle bisectors of the angles of a triangle meet at a point called
the incenter. The incenter is the center of the inscribed circle or incircle
of the trianglea circle that is tangent to all three segments making up the
sides and inside the triangle. The incenter of 4A00 B 00 C 00 in figure A.7 is I.
See also figure A.8. A Geometer proof of this fact can be found in the file
Teaching/ThreeIn.T.
Altitude. An altitude of a triangle is a line passing through one of its vertices and
perpendicular to the side opposite that vertex. The three altitudes of a triangle
meet at a point called the orthocenter. The orthocenter of 4A0 B 0 C 0 in figure A.7 is H. Be carefulalthough both the centroid and incenter are always
inside the triangle, the orthocenter may be outside it. A Geometer proof that the
three altitudes meet at a point can be found in Proofs/Orthocenter.T.
Orthocenter. The point where the three altitudes of a triangle intersect.
Excenter. If a triangle is considered to be three lines rather than three segments, there
are four circles tangent to all three lines. One, the incircle, is tangent to the three
segments. Each of the others, called an excircle, is tangent to the interior of
one segment, and to the other two lines exterior to the triangle. The center of
an excircle is called an excenter. Every triangle has three excircles and three

APPENDIX A. MATHEMATICS REVIEW

306
excenters. See Section A.6.8.

Circumcircle. The circle passing through all three vertices of a triangle. See figure A.8.
Circumcenter. The circumcenter of a triangle is the center of the circle (called the
circumcircle) that passes through the three vertices of the triangle. The center
of the circumcircle can be found at the intersection of the three perpendicular
bisectors of the segments that make up the sides of the triangle. See figure A.8.

A.6.1 Congruence and Similarity of Triangles

There are a few standard methods to prove that two triangles are congruent. In high
school textbooks, they are usually listed as an abbreviation like SAS, ASA, SSS,
or AAS. These stand for Side-Angle-Side, Angle-Side-Angle, et cetera.
SAS, or Side-Angle-Side stands
C
for the idea that if two of the sides in one
triangle have the same lengths as two sides
D
in another triangle, and further that the
angles between those two sides are equal
A
B
in both triangles, then the triangles themselves are congruent. Because they are
Figure A.9: SSA Failure
congruent, you then automatically know
Review/Ssa.D [D]
that all the measurements in the two triangles are the samethe other side and the
other two angles. (And in fact you also know that corresponding medians will be
equal, that corresponding altitudes will be equal, that the areas of the two triangles will
be equal, and so on. The order SAS is important, and can be interpreted as the order
that you read off the parts of the triangles as you go around. First there is a side, then an
angle, then another side. There is no SSA method because if the angle is not the included anglethe angle between the two sidesthen the triangles are not necessarily
congruent.
In figure A.9 are two triangles that are clearly not congruent, but they have two pairs
of equal sides and an equal angle, but not the angle between them. The two triangles
are 4ADB and 4ACB. Side AB is in both of them, so that is one match. Sides AD
and AC are also clearly equal since they both lie on a circle centered at A, and both
triangles share the angle at B.
Although SSA is not in general a valid method of proof of the congruence of
a pair of triangles, it is valid in the special case that the angle is at least 90 , and in
the special case where it is exactly 90 the method is known as the hypotenuse-leg
method. It is quite instructive to play around with the Geometer illustration for figure A.9 and try to adjust the vertices so that the angle at B is a right angle or larger and
see why in this case an SSA-where-A-is-a-right-angle-or-larger congruence theorem
is true.

A.6. TRIANGLES

307

ASA states that two triangles are congruent if two angles and the side between
them are congruent. Both ASA and AAS are equivalent for the simple reason that
if two of the angles of a triangle are equal to the two corresponding angles in another,
then the third angles of both must be equal as well, since the three angles of a triangle
always add to 180 .
Finally, if all three sides in a triangle are equal in length to the three sides of another
triangle, then the triangles are congruent, so the corresponding angles will also be
equal. This is known as the SSS, or Side-Side-Side method3 .
Proving the similarity of two triangles is almost always done by showing that they
have two equal angles (and hence, three equal angles). This is called Angle-Angle
similarity, or just AA.
There are also the SAS and SSS similarity conditions: If the included angles
between pairs of sides in two triangles are equal, and if the ratios of the lengths of the
sides are equal, then the triangles are similar. Finally, if all three sides of two triangles
are in the same ratio, the triangles are similar. In practice, these last two methods are
seldom used.

A.6.2 General Properties of Triangles

The lengths of any two sides of a triangle added together are longer than the
third side. In other words, if a triangles sides have lengths a, b, and c, then
a + b > c (and or course b + c > a and c + a > b). This is known as the triangle
inequality.

E
C

I
B

Figure A.10: Elementary Triangle Properties

Review/Triprop.D [D]

If a line is parallel to one side of a triangle, it divides the other two sides proportionally. In figure A.10, if DE k BC then AE : EC = AD : DB,
AE : AC = AD : AB, et cetera.
A line parallel to one side of a triangle cuts off another triangle that is similar to
the original triangle. In figure A.10 if DE k BC then 4ABC 4ADE.
3 The fact that three lengths determine a triangle is key to many mechanical construction techniques. If
three bars are bolted together to form a triangle, that triangle cannot change shapes unless the bars bend or
break. Geodesic domes are constructed entirely of triangles, and many rectangular features in architecture
are braced across the diagonal, essentially turning the flexible rectangle into two non-flexible triangles.

APPENDIX A. MATHEMATICS REVIEW

308

The median of a triangle divides the triangle into two triangles of equal area.
More generally, a line from a vertex of a triangle to the opposite side divides
the triangle into two triangles with areas proportional to the lengths into which
that side was divided. In figure A.10, F I : IH = A(4F IG) : A(4HIG), no
matter where I lies on the interior of the segment F H.
The angle bisector of a triangle divides the side opposite the bisected angle into
two segments proportional to the lengths of the adjacent sides of the original
triangle. In figure A.10, if F GI = IGH then F I : IH = GF : GH. A
Geometer proof appears in Proofs/Anglebisect1.T.
In 4ABC, if the lengths of the sides opposite the angles A, B, and C are a, b,
and c, respectively, then if A = B we know that a = b. Conversely, if a = b,
then A = B. Finally, if A > B, then a > b, and conversely.
In other words, in any triangle, the larger sides are opposite
larger angles. So in
the 30 60 90 triangle whose sides have lengths 1, 3 1.7320508,
and 2,

the side of length 1 is opposite the 30 angle, the side of length 3 is opposite
the 60 angle, and the side of length 2 is opposite the right angle.
This theorem is usually proved in the case of equality of the sides or angles by
dropping a perpendicular line to divide the triangle into two smaller ones that are
shown to be congruent by SAS or AAS.
Here is a beautiful proof that sides opposite equal angles are equal and viceversa. Suppose that in 4ABC, AB = AC. Then by SAS, 4BAC
= 4CAB,
so ABC = ACB. Similarly, if ABC = ACB, then 4BAC
= 4CAB
by ASA, so we can conclude that AB = AC. Both proofs show that a triangle
is congruent to the mirror-image of itself.

A.6.3 Area of a Triangle

The area of a triangle is half of the base times
the altitude. Remember that the altitude is the
perpendicular line to the base from the vertex
opposite the base. The altitude may not even
intersect the base inside the triangle. For any
triangle, there are three base-altitude pairs that
can be used to calculate the area, and all of
them give the same result, so one pair may be
easier to use than another.
Figure A.11 shows the three ways to calculate the area of 4ABC using the base-altitude
formula. The three altitudes are AA0 , BB 0 ,
and CC 0 . Notice that two of the altitudes,
AA0 and BB 0 lie outside the original triangle
4ABC, although in a general triangle they are

C
C

Figure A.11: Area of a Triangle

Review/Trianglearea.D [D]

A.6. TRIANGLES

309

sometimes in and sometimes out. Any of the following expressions gives the area of
4ABC:
BC AA0
CA BB 0
AB CC 0
=
=
.
A(4ABC) =
2
2
2
From the fact that the area of a rectangle is equal to its base times its height, it is
very easy to conclude that the area of a triangle is half the base times the height. In
figure A.12, suppose the problem is to determine the area of the triangle at A. Make an
exact copy of the triangle, but flip it over and place it against itself as in B. The area
of this parallelogram will be twice the area of the original triangle. To find the area of
the parallelogram, cut off the right triangle at the right end of the parallelogram, as in
C. Move that part to the left as in D. Thus the area of the rectangle at D is twice the
area of the original triangle. The base of the rectangle at D is the same as the base of
the original triangle, and it is also easy to see that the altitude of the rectangle is the
same as the altitude of the original triangle. The area of the rectangle is its base times
its height, which is the same as the base times height of the original triangle, but since
the rectangles area is made up of two copies of the triangle, the triangles area is half
that of the rectangle.
The argument in the paragraph is
intuitive, but is not quite rightif the
altitude hits the base very far outside
the segment which is the base, it may
not be possible to slice the paralleloA
B
gram as described. If the top point of
the triangle written A in figure A.12 is
moved to the right by a long distance,
the argument does not work. The argument can be made rigorous most easily
D
C
with an appeal to Cavalieris principle
which states that if two plane figures
can be placed so that all lines paralFigure A.12: Area of a Triangle
lel to a given line cut equal length segReview/Trianglearea1.D [D]
ments in both figures, the two figures
4
have equal area .

A.6.4 Angles in a Triangle

The three interior angles of a triangle add to 180 . Figure A.13 contains most of a proof
of this fact. In any 4ABC, find the midpoints of the sides (X, Y , and Z) and connect
them. It is easy to show that all four of the smaller triangles produced are congruent.
That means that the three angles that meet at Z: AZY , Y ZX, and XZB are
equal to the vertex angles in the original 4ABC. But they also fit together perfectly
on the line AB, so the sum of the three angles of any triangle is 180 . A Geometer
proof of this fact appears in Proofs/Sum3angles.T.
4 The three-dimensional version of Cavalieris principle states that if two solids can be arranged so that
all planes parallel to a given plane intersect the solids with equal areas, the solids have equal volumes.

APPENDIX A. MATHEMATICS REVIEW

310

C
X
B

Y
Z

Figure A.13: Sum of Angles in a Triangle

Review/Sum3angles.T [P]

A trivial (but often very useful)

consequence of this is that any external angle of a triangle is equal to the
sum of the other two interior angles.
Figure A.13 shows this as well. In
4BXZ, the exterior angle at Z is
AZX, and that is clearly equal to the
sum AZY + Y ZX which are equal
to the other two angles in the original
4BXZ (remember that all four of the

smaller triangles in this figure are congruent).

In a right triangle, of course, this means that the two acute angles must add to make
a right angle (in other words, they are complementary) since the exterior angle of the
right angle is also a right angle.

 The theorem above cannot be proved without the so-called parallel postulate
which states that given a line and a point not on the line, there exists exactly one
line passing through that point that is parallel to the given line. In other geometries,
Riemannian geometry, projective geometry, and others, the parallel postulate does not
hold, and the angles of triangles in those geometries need not add to 180 .
 To see why it might be reasonable to consider a geometry where a triangles
angles do not add to 180 , imagine that you live on a sphere (that should not be too
hard) and that you measure angles of triangles on its surface. If the triangles are small,
the angles will add to almost 180 , but as they get larger, the error becomes greater.
Consider a trip along the triangle that begins at
A
the north pole and goes due south along the prime

meridian (0 longitude) through the observatory at

c b
Greenwich, England, then through Spain near VaC
Ba
lencia, then through north Africa, finally exiting that
continent through Ghana. Just after you enter the
Atlantic Ocean, you will arrive at the equator, where
you then take a 90 turn to the west and go a quarter Figure A.14: Pythagorean Theof the way around the earth through South America orem
and entering the Pacific Ocean from Ecuador. After
Review/Pythagorean.D [D]
you plunge into the Pacific a bit north of Guayaquil,
continue a few hundred kilometers along the equator to the 90 line of latitude somewhere near the middle of the Galapagos Islands, and finally take another 90 turn due
north passing near Guatemala City, through the Yucatan in Mexico, and very near New
Orleans, Louisiana. Continue along this line all the way to the north pole where you
close the triangle. All the paths you took were as straight as you could make them, and
all three angles in that giant spherical triangle measure 90 , so the total of the interior
angle measurements is 270. If you are willing to look at even larger triangles that
cover more than 1/8 of the globe, you can make the total measure as close to 540 as
you want. For example, if you go past the Galapagos Islands and across the Pacific to
near the international date line before you turn north, you can make the angle at the
north pole as close to 180 as you would like.

A.6. TRIANGLES

311

A.6.5 The Pythagorean Theorem

Certainly the most famous, and probably the most useful theorem in all of Euclidean
geometry is the pythagorean theorem. Not only is it useful in standard geometric
proofs, but it is also the basis for coordinate geometry, or so-called analytic geometry (see Section A.11).
Here is a statement of the theorem:
Theorem A.1 (Pythagorean Theorem) The square of the length of the hypotenuse of
a right triangle is the sum of the squares of the lengths of the opposite two sides and
conversely, if the sum of the squares of two sides of a triangle is equal to the square of
the third side, the triangle is a right triangle.
C

Figure A.15: The Pythagorean Theorem Proof

Review/Pyproof1.D [D]

The theorem is illustrated in figure A.14.

In 4ABC, where C = 90 , squares are
constructed on each of the sides. The length
of sides AB, BC, and CA are c, a, and b, respectively. Thus the squares have areas of a2 ,
b2 , and c2 . The pythagorean theorem simply
states that a2 + b2 = c2 . There are literally
hundreds of proofs of this theorem; what follows are two proofs, the first of which is computational, and the second is more visual.

In figure A.15, 4ABC is a right triangle

with C = 90 . Drop an altitude from C
that meets the segment AB at point D. Because of the AA similarity theorem, all three
of the triangles in the figure are similar: 4ABC 4CBD 4ACD.

By similarity, we have AD/AC = AC/AB and DB/CB = CB/AB. Multiply

these out, giving: AC 2 = AB AD and CB 2 = AB DB. Add them: AC 2 + CB 2 =
AB DB + AB AD = AB (DB + AD) = AB AB = AB 2 , which completes the
proof.
Here is another proof which combines a nice visual and analytic approach:
b

a
b

c
b

Figure A.16: Proof 2 of the Pythagorean Theorem

Review/Pyproof2.D, Review/Pyproof3.D [M]

APPENDIX A. MATHEMATICS REVIEW

312

Make four copies of the right triangle whose sides have lengths a, b, and c and
arrange them to form a square as shown in the left of figure A.16.
We can calculate the area of the square in two ways. If we just square the length
of a side, we get (a + b)2 . If we notice that the square is composed of a central square
and 4 equal triangles, we get: c2 + 4(ab/2) = c2 + 2ab. Setting those areas equal and
simplifying, we obtain:
(a + b)2
a2 + 2ab + b2
a2 + b 2

= c2 + 2ab
= c2 + 2ab
= c2 .

But there is even a more visual proof. Take the squares and triangles on the left and
rearrange them as in the diagram on the right in figure A.16. Both figures have four
triangles, and the remaining area is either two squares with areas a 2 and b2 or one with
area c2 . Clearly a2 + b2 = c2 .
We obviously need to show that both the figures above are, indeed, squares and that
the object in the center in the right side of figure A.16 is also a square, but this is easy
since we know that the three angles of a triangle add to 180 and that in a right triangle,
one of the angles is 90 .

A.6.6 Stewarts Theorem

Although this theorem does not normally appear in high school geometry courses, it is
easy to prove and it is extremely useful:

C
b
A m

D
c

A m

t
c

E n-t

Figure A.17: Stewarts Theorem

Review/Stewart.D [D]

Theorem A.2 (Stewarts Theorem) In any triangle 4ABC, let D be any point on
the segment AB. If the segment lengths are: a = BC, b = AC, c = AB, d = CD,
m = AD and n = BD then:
a2 m + b2 n = c(mn + d2 ).
The proof is fairly simple. Drop a perpendicular line from C to AB and suppose
it intersects at E between D and B as in the diagram on the right in figure A.17. (A

A.6. TRIANGLES

313

complete proof requires that all possibilities of the position of E be considered, but all
are similar.) Then the pythagorean theorem yields:
d2

= t2 + h2

a2

= (n t)2 + h2

b2

= (t + m)2 + h2

As direct results of these equations and remembering that c = m + n we obtain:

a2 m + b2 n = m((n t)2 + h2 ) + n((t + m)2 + h2 )
c(mn + d2 ) = (m + n)(mn + h2 + t2 )

(A.1)
(A.2)

The right sides of equations A.1 and A.2 are readily shown to be equal, so the proof
is complete.
Stewarts theorem can also be proved by applying the law of cosines repeatedly.
An interesting special case of Stewarts theorem occurs if the angle at C is double
the angle at A. In this case c2 = a(a + b).

A.6.7 Special Triangles

Some triangles appear over and over because they are useful, either in a practical sense,
or because they serve as good examples or as examples that are easy to use for sample
calculations. One set of them is interesting because the angles are nice round numbers,
and the other set because the lengths of the sides are nice round numbers. (Of course it
is easy to make up triangles where the angles are nice round numbers, but for the vast
majority of them, it is impossible to calculate analytically the lengths of the sides and
vice-versa.)

Figure A.18: Special Triangles

Review/Special.D [D]

Triangles with nice angles include the 60 60 60 triangle, the 30 60 90

triangle, and the 45 45 90 triangle. See figure A.18. The first one, (4ABC) is
equilateralall three sides and angles are the same. The second (4DEF ) is just the
equilateral triangle sliced in halfits three edges have lengths in the ratio EF : F D :

APPENDIX A. MATHEMATICS REVIEW

314

DE = 1 : 2 : 3. The 45 45
90 triangle (4GHI) is isosceles, and the ratios of
the lengths of
its sides are 1 : 1 : 2. Remember that the larger angles are opposite the
longer sides ( 2 1.41421 and 3 1.73205).

The other triangles you see regularly have sides whose lengths have integer ratios
and at the same time are right triangles. Any three integers a, b, and h will form a right
triangle with h as the hypotenuse if a2 +b2 = h2 (this is a result of the pythagorean theorem). Here are a few examples of so-called pythagorean triplets: (3, 4, 5), (5, 12, 13),
(6, 8, 10) (of course the (6, 8, 10) is similar to the (3, 4, 5) trianglejust twice as big),
and (7, 24, 25). A reduced pythagorean triplet is one where the entries have no
factors in common. (6, 8, 10) is not reduced because there is a factor of 2 in every
number.
The pythagorean triples do not have reasonable angles. For the (3, 4, 5) triangle,
for example, the acute angles are approximately 36.869898 and 53.130102.
 There are an infinte number of reduced pythagorean triplets. If m and n are any
two integers with m > n > 0, then the following three numbers form a pythagorean
triplet: (m2 n2 , 2mn, m2 + n2 ). It is easy to show that any such set is a triplet, but
it is also true that all pythagorean triplets (reduced and non-reduced) are of this form.
If one of m or n is even and if they are relatively prime, then the formula will produce
a reduced pythagorean triple.
 It is not hard to prove that all pythagorean triples have this form. The problem
of finding points with rational coordinates on the unit circle x2 + y 2 = 1 is equivalent
to finding integral pythagorean triples. Every line that passes through a point with
rational coordinates and the point A = (1, 0) has a rational slope, and vice-versa. The
equation of the line passing through A and having slope n/m is y = (n/m)(x 1)
and if we plug that into the equation of the circle we find that the rational point of
intersection other than A is at ((m2 n2 )/(m2 + n2 ), 2mn/(m2 + n2 )) which is
equivalent to the solution in the previous paragraph.
Table A.1 lists a few smaller examples of the form: (n, m) (2mn, m2
n , m2 + n2 ). Examples followed by an asterisk (*) are reduced pythagorean triplets.
2

(1, 2)
(1, 5)
(2, 3)
(2, 6)
(3, 4)
(3, 7)
(4, 5)
(4, 8)

(4, 3, 5)
(10, 24, 26)
(12, 5, 13)
(24, 32, 40)
(24, 7, 25)
(42, 40, 58)
(40, 9, 41)
(64, 48, 80)

(1, 3)
(1, 6)
(2, 4)
(2, 7)
(3, 5)
(3, 8)
(4, 6)
(4, 9)

(6, 8, 10)
(12, 35, 37)
(16, 12, 20)
(28, 45, 53)
(30, 16, 34)
(48, 55, 73)
(48, 20, 52)
(72, 65, 97)

(1, 4) (8, 15, 17)

(1, 7) (14, 48, 50)
(2, 5) (20, 21, 29)
(2, 8) (32, 60, 68)
(3, 6) (36, 27, 45)
(3, 9) (54, 72, 90)
(4, 7) (56, 33, 65)
(4, 10) (80, 84, 116)

Table A.1: Some Pythagorean Triples

Students taking standardized tests in geometry, or participating in high school
mathematics contests, should commit a few of the smaller examples above to memory because they appear all the time! In a recent AHSME high school math contest

A.6. TRIANGLES

315

(American High School Mathematics Examination for 1999) one of the questions concerned a circle circumscribed about a triangle with sides 20, 21, and 29. Those who
knew that the (20, 21, 29) triangle is a right triangle knew that the side of length 29
must be the diameter of the circle, and the problem was trivial5 .

A.6.8 Inscribed and Escribed Circles

A
s-a

s-a

s-a
s-b

s-c
C s-c

s-b

s-a
B

s-b
s-c

s-c

A s-b
s-c
s-a
s-a

s-c

s-b
C
s-b

Figure A.19: Lengths to Points of Tangency

Review/Trilengths.D [D]

Every triangle has four circles tangent to its three sidesthe incircle and three
excircles. There are a number of easily derived lengths in these triangles that are extremely useful for calculations. In figure A.19 above, a, b, and c are the lengths of
the sides opposite vertices A, B, and C, respectively, and s = (a + b + c)/2 is the
semiperimeter (half the perimeter). The lengths marked s a, s b and s c are
measured from the vertex of the triangle to the points of tangency of the inscribed or
escribed circle. The figure includes two triangles 4ABC since otherwise it would be
too cluttered to understand. If all four tangent circles were drawn in one triangle, each
length, s a, s b and s c, would measure six different segments.
These are not hard to derive. If we do not know the lengths, call the lengths from
A, B and C to the points of tangency with the incircle , and , respectively. The
perimeter of 4ABC = 2 + 2 + 2 = a + b + c = 2s.
But we also know that + = c, + = a and + = b. A little algebra shows
that = s a, = s b and = s c.
A similar observation gives us the lengths from the vertices to the excircles. In the
figure on the right, let us temporarily call the sides labeled s a, s b and s c ,

5 For those participating in math contests, another interesting triangle to know about is the one whose
sides are 13, 14, and 15. It is just a 5-12-13 triangle back to back with a 9-12-15 triangle. The 9-12-15
triangle is the 3-4-5 triangle in disguise.

APPENDIX A. MATHEMATICS REVIEW

316

and , respectively. Since the distances from A to the tangent points with the excircle
on the other side of BC are the same, we have c + = b + . We also see that
+ = a = 2s b c. If we solve these two equations for , we obtain = s b.
Exactly the same derivation gives = s a and = s c.
There is more to say, however. Let r be the radius of the incircle. Then the area of
4ABC, A(4ABC) = r(a + b + c)/2 = rs. This is because if I is the incenter of
4ABC and we split 4ABC into 4AIB, 4BIC and 4CIA. Those three triangles
have bases of a, b and c and all have altitude r.

A similar analysis shows that if ra , rb and rc are the radii of the excircles opposite
vertices A, B and C, we have:
A(4ABC) = ra (s a) = rb (s b) = rc (s c) = rs.

(A.3)

A little more algebra yields the following relationship:

1
1
1
1
+
+
= .
ra
rb
rc
r

(A.4)

A.7 Quadrilaterals
The term quadrilateral is the general name for a polygon having four sides. The situation becomes surprisingly more complicated when one goes from three-sided triangles
to four-sided quadrilaterals, so there are many special names and definitions associated
with quadrilaterals. With more than four sides, the situation is so complicated that
mathematicians have basically given upthere are very few names for special polygons other than the regular ones for polygons with five or more sides. A quadrilateral
is just a special case of a polygon, so remember that general polygon definitions (see
Section A.8) apply to quadrilaterals (and even triangles) as well.
For triangles there is a special triangle symbol for notation, so the triangle with vertices A, B, and C is written 4ABC. Quadrilaterals have so many shapes that it would
be somewhat misleading to write something like ABCD for the quadrilateral with
vertices A, B, C, and D (although many texts do). Here we will simply say something
like, the quadrilateral ABCD or if the specific shape is known, the square ABCD
or the trapezoid ABCD.
Here are some specific definitions related to quadrilaterals.
Square A square has all four sides and all four angles equal. Of course, the angles
are all right angles. A square is a regular quadrilateral. See A in figure A.20.
Rectangle A rectangle has four right angles, but not necessarily four equal sides. See
A and B in figure A.20.
Parallelogram A parallelogram has as its edges two pairs of parallel line segments.
See A, B, C, and D in figure A.20.
Rhombus A rhombus has all four sides equal. See A and D in figure A.20.

A.7. QUADRILATERALS

317

Figure A.20: Special Quadrilaterals

Review/Quadrilaterals.D [D]

Trapezoid A trapezoid has two sides parallel, but not necessarily the other two. See
anything except F in figure A.20.
Isosceles Trapezoid An isosceles trapezoid is a trapezoid that is not a parallelogram, and the two non-parallel sides have the same length.
Kite A kite is a convex polygon both of whose pairs of adjacent sides are of equal
length. See A, D, and F in figure A.20.

A
H

F
G

D
Figure A.21: Inscribed and Circumscribed Quadrilaterals
Review/Circquads.D [D]

Cyclic Quadrilateral If there exists a circle such that the quadrilateral (or any polygon, for that matter) has all its vertices on it, the quadrilateral is called a cyclic
quadrilateral or a concyclic quadrilateral. Quadrilateral EF GH is inscribed
in the circle and hence is a cyclic quadrilateral. See figure A.21.
Circumscribable Quadrilateral Similarly, if there is a circle that is tangent to all four
sides of a convex quadrilateral, it is called a circumscribable quadrilateral. In
figure A.21, quadrilateral ABCD is circumscribed about the circle.
Convex Quadrilateral A convex quadrilateral is a special case of a convex polygon
(see Section A.8). A shape is convex if a line segment connecting any two interior points is completely inside the shape. All the quadrilaterals in figure A.20 are
convex. The quadrilateral EF GH in figure A.22 is not. A non-convex quadrilateral (or any non-convex shape) is sometimes called concave.

APPENDIX A. MATHEMATICS REVIEW

318

C
E
B

Figure A.22: Bowtie and Non-Convex Quadrilateral

Review/Badquads.D [D]

Bowtie Quadrilateral A bowtie quadrilateral (or non-simple quadrilateral) is a

quadrilateral whose edges cross each other such as ABCD in figure A.22. This
cannot happen with triangles, but once a polygon has more than three sides,
this is possible, and it is the bane of computer graphics programmers. When
the edges of a polygon cross themselves, it no longer makes sense to talk about
inside and outside.

A.7.1 Quadrilateral Theorems

Many properties of quadrilaterals are simple consequences of the fact that two opposite
vertices can be connected by a segment dividing the quadrilateral into two triangles
whose properties are easy to calculate. For anything other than special quadrilaterals
like rectangles or concyclic quadrilaterals, usually the easiest way to find the area, for
example, is to divide it into two triangles and find the sum of the areas of those two
triangles.
However, here are a few important facts about quadrilaterals, in no particular order.
The vast majority are easy to prove:
The four internal angles of a non-bowtie quadrilateral add to 360 . For bowtie
quadrilaterals, there is no such thing as an internal angle.
The diagonals of a rectangle are congruent.
The diagonals of a rhombus are perpendicular and conversely if the diagonals of a parallelogram are perpendicular, the parallelogram is a rhombus. See
Proofs/Rhombus.T for a Geometer proof.
The diagonals of a rhombus bisect the angles at the vertices of the rhombus.
The opposite sides of a parallelogram are of equal length and the opposite angles
have equal measures. See Proofs/EquParSides.T for a Geometer proof.
Any pair of adjacent angles in a parallelogram are supplementary.
Either diagonal of a parallelogram divides it into two congruent triangles. See
Proofs/EquParSides.T for a Geometer proof.

A.8. GENERAL POLYGONS

319

The diagonals of a parallelogram bisect each other. See Proofs/ParBisect.T for

a Geometer proof.
If a parallelogram has sides of lengths a and b and diagonals of lengths x and y,
then 2a2 + 2b2 = x2 + y 2 .
The area of a parallelogram is the base times the height. Warning: the height
is the perpendicular distance from the base to the side parallel to itit is not the
distance along the other edge.
If the midpoints of the edges of any quadrilateral are connected in order, the resulting figure is a parallelogram. This remains true even for bowtie quadrilaterals. As long as it is not a bowtie quadrilateral, then the area of the parallelogram
is half the area of the quadrilateral. This is known as Varignons theorem.
An isosceles trapezoid can be inscribed in a circle. See Proofs/Isotrap.T for a
Geometer proof.
Isosceles trapezoids have congruent base angles. See Proofs/IsoTrapAng.T for
a Geometer proof.
The area of a trapezoid is the average of the lengths of the parallel sides times
the distance between the sides.
If a quadrilateral ABCD can be circumscribed about a circle, then AB + CD =
AD + BC and conversely. See Proofs/QuadEdges.T for a Geometer proof.
If a convex quadrilateral ABCD can be inscribed in a circle, then A + C =
B + D = 180 and conversely.

A.8 General Polygons

Given n different points in the plane, A1 , A2 , . . . , An , a polygon is the set of segments
A1 A2 , A2 A3 , A3 A4 , . . . , An1 An , and An A1 . A polygon has three or more sidesit
is usually not very useful to think of one-sided or two-sided polygons since then your
theorems usually would have the form, For all polygons except those with one or two
sides . . . . There is no upper limit, of course, to the number of sides a polygon may
have. The points A1 , . . . , An are called the vertices of the polygon, and the segments
A1 A2 , . . . , An A1 are called the edges. Sometimes it is important to talk about the
angles of the polygon which are A1 A2 A3 , A2 A3 A4 , . . . , An A1 A2 . Notice that
this is a quite general definition of a polygon and allows the edges to cross each other,
et cetera6 . Special names exist for more restricted classes of polygons. There are a
couple of different schemes for naming polygons (including the names for 1- and 2sided polygons), and the more common ones are listed in Table A.2. The names used
in this book are listed first, and one alternative is listed in parentheses.
6 There are, however, still more general definitions of polygons that allow holes or which allow polygons
to have more than one component. Although such definitions are sometimes useful in computer graphics,
they will not be considered here.

APPENDIX A. MATHEMATICS REVIEW

320
1
2
3
4
5
6
7
8
9
10

monogon
digon
triangle (trigon)
quadrilateral (tetragon)
pentagon
hexagon
heptagon
octagon
enneagon
decagon

11
12
13
14
15
16
17
18
19
20

hendecagon (undecagon)
dodecagon
tridecagon (triskaidecagon)
tetradecagon (tetrakaidecagon)
pentadecagon (pentakaidecagon)
hexadecagon (hexakaidecagon)
heptadecagon (heptakaidecagon)
octadecagon (octakaidecagon)
enneadecagon (enneakaidecagon)
icosagon

Table A.2: Names of polygons

Here are some definitions relating to general polygons. Remember that triangles
and quadrilaterals are special cases of polygons, so we can talk about the perimeter of
a triangle or square, for example.
Perimeter The perimeter of a polygon is the sum of the lengths of the edges of the
polygon. It is the distance around the polygon.
Simple Polygon A simple polygon is one whose non-adjacent edges do not meet,
even at the endpoints. In other words, there are no crossing lines, or places
where the polygon comes back and touches itself.
 Although it is obvious what is meant by the inside and outside of a simple
polygon, in practice (for example, when you are trying to write a computer program that deals with general polygons) it can be difficult to determine whether a
point is inside or outside a given polygon, especially if the polygon has hundreds
or thousands of sides twisted up like a maze. The usual method is based on the
Jordan curve theorem which states that any simple closed curve (simple means
that it does not cross itself, and a sequence of straight edges is just a special kind
of curve) divides the plane into two regionsan inside and an outsidewhere
the inside has finite area and the outside infinite area. Furthermore, if a point
on the inside is connected to a point on the outside by a straight line, that line
must intersect the curve. (Although it is obvious, this is not an easy theorem
to prove.)
From the Jordan curve theorem, one can conclude that if you are given a point
in the plane not on some simple polygon and you wish to decide if it is inside
or outside, draw a ray from that point out to infinity in some arbitrary direction,
and count the number of times it crosses the curve (or in your case, segments of
the polygon). If the ray crosses an odd number of times, the point was inside;
otherwise, it was outside. Even this can be tricky to do in a computational sense.
The ray may just touch a vertex of the polygon (this does not count as a crossing),
or it may be coincident with an edge for a while. In this second case, whether it
crosses or not depends on the relative directions that the segments connected to
the endpoints of the coincident segment take.

A.8. GENERAL POLYGONS

321

Convex Polygon A convex polygon is a polygon such that given any two points in
the interior, the line segment connecting them will also be completely contained
in the interior. In other words, a convex polygon does not have any indentations.
A star-shaped polygon, for example, is not convex.7 Every triangle is convex,
but if it has more than three sides, a polygon may be non-convex.
Regular Polygon A regular polygon is a simple polygon having any number of sides,
but all of those sides have the same length, and all the internal angles are the
same. We have special names for the smaller onesan equilateral triangle and
a square in the case of three and four sidesbut after that, we simply say a
regular pentagon, a regular hexagon, and so on.
Triangulation A triangulation is a subdivision of a polygon into triangles. Any simple polygon can be triangulated using only segments that connect existing vertices of the polygon. Although this may seem obvious, it is not easy to prove
see Section A.8.1.1.

A.8.1 General Polygon Theorems

At first glance, it seems that an unreasonable amount of time is spent studying the
properties of a very special type of polygonthe triangle. After all, why not spend an
equivalent amount of time studying 4-sided, 5-sided, and 6-sided figures?
A.8.1.1 Polygon Triangulation
The reason is that triangles are relatively simple, and that most other interesting planar
figures bounded by straight lines can be decomposed into triangles. If you know everything about triangles, you know a great deal about more complicated geometric figures
composed of line segments.
On the left side of figure A.23, for example, we see a polygon that is quite complicated and has 17 sides.
However, on the right side of the same figure, we see that it is possible to break it
up into 15 different triangles: 4ABQ, 4BQE, 4DBE, et cetera. If you know all
the details of the little triangles composing it, you also know many of the details of
the more complex polygon. For example, if you would like to know the measure of
CDE in the original polygon, you can look at 4CBD and 4BDE, and find the
angles CDB and BDE. Those can then be added to get the desired angle in the
general polygon.
A.8.1.2  Triangulation Techniques
Notice that in this example, the subdivision of the complex polygon into triangles was
done without adding any vertices, even though the original polygon had all sorts of
7 The

word convex can be used to describe any figure with this propertynot just polygons.

APPENDIX A. MATHEMATICS REVIEW

322
G

H
F

H
F

O
J

O
J

C
E
A
Q

P
M L

I
K

A
Q

P
M L

I
K

Figure A.23: Triangulation of a Polygon

Review/Polygon.D [M]

weird indentations. It is always possible to do this without adding any additional vertices. One proof is beyond the scope of this book, but it involves looking for ears
sections of the polygon where a triangle sticks out and can be cut off. If there is always
an ear available, clearly any polygon can be so subdivided. Simply find an ear, cut it
off, and you are left with one triangle and a polygon with one side fewer. Find the ear
on that one, cut it off, and continue the process until all that is left is a triangle.
Another approach to show that an arbitrary polygon can be triangulated finds an
interior diagonal (a line connecting two vertices that is completely contained within
the polygon). This splits the polygon into two smaller ones, and the triangulation of
the polygon can be inferred by an inductive argument.
To prove the existence of an interior
G
diagonal, choose an angle like ABC
of polygon ABCDEF G in figure A.24
where the interior of the polygon is on
the side of the angle measuring less than
180 . If the segment AC lies comF
pletely within the polygon, then it is
A E
C
D
the required interior diagonal and we are
done. If not, there must be some finite
B
number of vertices in the polygon that
lie within 4ABC. For each such vertex, construct a line parallel to AC and
Figure A.24: Find an Internal Diagonal
choose a vertex whose line is closest to
Review/CutPoly.T [M]
B. That vertex can be connected to B
completely within the polygon since if it crossed another line, that other line would
have to be even nearer B. Note in figure A.24 that we cannot use the vertex closest to
Bthat would be vertex F . It is quite instructive to play with the Geometer version

A.9. CIRCLES

323

of the figure and experiment with various shapes of the polygon.

A.8.1.3 Total Interior Angle of a Polygon
Since every polygon can be divided into n 2 triangles, it is easy to see that the sum of
the interior angles of an n-sided polygon is (n 2) 180 . Every interior angle will be
composed of a sum of angles of the triangles, and all of the angles of all the triangles
will be used.
A.8.1.4 Area of a Polygon
As was stated before, sometimes the easiest way to find the area of a polygon is to
split it into triangles and then sum the areas of those triangles. If the vertices of a
polygon are described by coordinates in a cartesian coordinate system, however, there
is a simple formula to calculate the area. See Section A.11.1.

A.9 Circles
A circle is the set of all points that are equidistant from a given point. The given
point is called the center of the circle and the distance from the center to any point
on the circle is called the radius of the circle. As with polygons, only the points on
the boundary are part of the circle; the points inside (including the center) are not. The
name disk is often used if it is necessary to refer to the points on a circle together with
those inside.
Diameter The diameter of a circle is twice the radiusit is the distance across the
circle through the center.
Circumference The circumference of a circle is the distance around the outside of
the circle. It is equal to 2r or d, where r is the radius or d is the diameter of
the circle. is just a number, approximately equal to 3.14159265.

D
O

E
C

Figure A.25: Circle Features

Review/Circledefs.D [D]

Tangent A line and a circle can have zero, one, or two points in common. If there is
one point in common, we say that the line is tangent to the circle and the line is
called the tangent line. In figure A.25, the line passing through T is tangent to

APPENDIX A. MATHEMATICS REVIEW

324

the circle. In a similar way, two distinct circles can have zero, one, or two points
in common, and if there is only one, the two circles are said to be tangent at
that common point.
Chord If a line intersects a circle at two points, then the segment of the line inside and
on the circle is called a chord. The entire line that cuts a circle in two places is
called a secant. The segment AB is a chord, and the line DC is a secant of the
circle in figure A.25.


Arc A connected section of a circle is called an arc. Note that if points A and B are
on a circle, then saying arc AB (denoted by AB) is ambiguousthe points
determine two arcs, depending on which way around you go.


Except in the case where the two arc endpoints are diametrically opposite, one
of the arcs is larger than the other, and AB usually refers to the smaller of
the two. Another reasonable convention is to label the points so the arc goes
counterclockwise from the first to the second. But if a theorem simply talks
about two arbitrary points on a circle and the arc between them, usually the
theorem only makes sense when the shorter one is meant. In this book, we will
label arcs using the counterclockwise method when there is an accompanying
diagram with a particular configuration.
Sometimes a third point can be given that lies between the endpoints and when
the arc is referred to with three letters, it is the part of the circle containing that
intermediate point.
If both endpoints of an arc are connected to the center of a circle with a pair of
radii, the angle between the radii is the measure of the arc.
Central Angle An angle whose vertex is at the center of a circle is called a central
angle. An angle inside a circle whose vertex is on the circle is called an inscribed angle. AOD is a central angle and EDC is an inscribed angle in
the circle in figure A.25.

A.9.1 Circle Theorems

Here are the most important properties of circles:
The area of a circle is r 2 , where r is the radius of the circle. The symbol (the
Greek letter pi) is defined to be the ratio of the circumference of a circle to
its diameter. 3.1415926535898. Figure A.26 shows why this is reasonable
to believe. The circle is first divided into a set of pie-shaped segments which
are then rearranged to make something that is close to a rectangle. If the circle
were divided into more such segments, the approximation to the rectangle would
be better. Since the circumference of the circle is 2r, where r is its radius,
it is clear from the figures that the approximated rectangle will have one side
approximately equal to r and the other approximately equal to r(remember
that what was the circumference now appears as a bumpy top and bottom of the
rectangle approximation), so its area is approximately r 2 , as desired.

A.9. CIRCLES

325

Figure A.26: Circle Segment Rearrangement

Review/CircAreaFig.T [S]

The perpendicular bisector of a chord of a circle passes through the center of the
circle.
Two tangents to a circle from an external point are of equal length.
A line tangent to a circle at a point T is perpendicular to the radius that connects
T with the center of the circle and conversely, if a line is perpendicular to a radius
at the point the radius touches the circle, that line is tangent to the circle.

O
X

Figure A.27: Inscribed and Central Angles

Review/Circprops1.D [D]

If a central angle and an inscribed angle in a circle subtend the same arc, then
the inscribed angle is half the size of the central angle. As a trivial consequence,
any two inscribed angles that subtend the same arc have the same measure. In
figure A.27, AOB = 2AXB = 2AY B. See Proofs/InscribedAngle.T
for a Geometer proof.
Another trivial, but extremely important consequence of the item above is that
any angle inscribed in a half circle (a semicircle) is a right angle. Conversely, if
the hypotenuse of a right triangle is the diameter of a circle, then the apex of the
angle lies on the circle. In figure A.28 all three of the angles AXB, AY B,
and AZB are right angles since they are inscribed in a semicircle (the segment
AB is the diameter). This is known as Thales theorem.

APPENDIX A. MATHEMATICS REVIEW

326

Z
B

Figure A.28: Inscribed and Central Angles

Review/Circprops2.D [D]

P
A
D

A
C
D

Figure A.29: Intersecting Secants

Review/Circprops3.D [D]

If a secant cuts a circle at points A and B and another secant cuts the same circle
at points C and D, and if the two secants intersect at a point P , then P A P B =
P C P D. The point P can be inside or outside the circle. Figure A.29 shows
both situations with P , the intersection of secants AB and CD, lying either
inside or outside the circle.
In figure A.29 on the right, 2AP C = DB AC. In other words, the angle is
half the difference of the measures of the subtended arcs. A Geometer proof of
this can be found in Proofs/CircleProperties.T.

B
A
P

Figure A.30: Tangent and Secant

Review/Circprops4.D [D]

If a secant passing through a circle at points A and B intersects at point P a line

that is tangent to the circle at point T , then P A P B = P T 2 . (This is about
the same as the previous item, where the points C and D happen to be the same
point.) See figure A.30.
If a line is tangent to a circle at T , and T A is any chord of the circle, then the
measure of the arc cut off by the chord is twice the measure of the angle made
by the two chords. In other words, the central angle AOT in figure A.31 is
twice the angle AT B at the tangent: AOT = 2AT B. A Geometer proof
appears in Proofs/TangAngle.T.

A.10. TRIGONOMETRIC DEFINITIONS

327

Figure A.31: Tangent and Chord

Review/Circprops5.D [D]

A.10 Trigonometric Definitions

The word trigonometry, of course, means triangle measurement. Its practical applications are related to making actual measurementssurveyors, for example, can
measure some angles and distances and from those compute others. Trigonometry is
primarily concerned with a few mathematical functions that are defined by the lengths
and angles in right triangles.
A right triangle contains a right angle, so the other two angles add to 90 . That
means any two right triangles that share another angle are similar by AA, giving us all
the similarity relationships that relate the lengths of the sides. In a right triangle, if you
know one of the other angles and the length of any side, you can determine all the the
other angles and lengths.
As a concrete example let us consider all right triangles that also contain an angle
. The ratios of the side opposite the angle and the hypotenuse will be the same,
independent of the size of the triangle. In trigonometry, that particular ratio is given
the name sin , where is the angle. The particular example of = 30 will always
have a hypotenuse twice as long as the side opposite the 30 angle, so sin 30 = 1/2.
The other common trigonometric functions are given by other ratios of lengths of
sides in any right triangle that contains the given angle.
At first it seems quite limiting to consider only right triangles but that is not the
case. Not only is it possible to convert any triangle to a pair of right triangles by
dropping the appropriate altitude, but the standard trigonometric functions appear in
all sorts of other formulas. There are many more trigonometric relations than are listed
in this section.

A.10.1 Definitions of the Trigonometric Functions

In figure A.32, C is a right angle. The definitions of all the trigonometric ratios for the
angle BAC = in terms of the lengths of the sides a, b, and c appear in Table A.3.
In old texts one sometimes encounters a few other trigonometric functions that are
easily expressed in terms of the basic ones shown in Table A.3. These include the
versine (vers = 1 cos = 2 sin2 (/2)), the coversine (covers = 1 sin ),
the haversine (hav = (vers )/2 = (1 cos )/2 = sin2 (/2)), and the exsecant

APPENDIX A. MATHEMATICS REVIEW

328

B
a

Figure A.32: Trigonometric Ratios

Review/Trig.D [D]

sine:
cosine:
tangent:

sin = a/c cosecant:

cos = b/c secant:
tan = a/b cotangent:

csc = c/a = 1/ sin

sec = c/b = 1/ cos
cot = b/a = 1/ tan

Table A.3: Definitions of the Basic Trigonometric Functions

(exsec = sec 1). It is an interesting exercise to try to figure out why these
definitions might be important and to determine trigonometric relations among them.
For many practical applications, the definitions in Table A.3 are sufficient, although
the drawing only makes sense for the trigonometric functions of angles that are between 0 and 90 . For completeness, here is the official mathematical definition of the
trigonometric functions for any angle. The official definition also has the advantage
that it is much easier to work with.

(0,1)
cos
B

sin

sin
(-1,0)

A = (x,y) = (cos ,sin )

cos

I = (1,0)
sin
cos

(0,-1)

Figure A.33: Trigonometric Function Definitions

Review/Trig1.D [D]

In figure A.33, we see the so-called unit circle drawn in the standard x-y coordi-

A.10. TRIGONOMETRIC DEFINITIONS

329

nate system. The unit circle is centered at the origin (0, 0), and has radius 1. Three
examples of angles are shown in the figure, each centered at the origin, and each beginning on the positive x-axis. Angle is the one marked with three concentric rings;
angle includes angle and in addition the angle marked with two concentric rings,
and angle includes all of and the additional angle marked with one ring. In the
notation we usually use, angles , , and are equal to IOA, IOB, and IOC,
respectively, where the points A, B, and C are on the unit circle, and the angles are
measured counterclockwise from I, so IOC looks as if it is roughly 315 .
Since the signs of the trigonometric functions change only when an angle changes
from one quadrant to another, it is often convenient to give the quadrants names, and
Roman numerals are used. The quadrant where x > 0 and y > 0 is called quadrant I,
and the next three, in counter-clockwise order, are quadrants II, III and IV.
Every point on the unit circle determines an angle, and the trigonometric functions
are simply defined in terms of the coordinates of those points. For example, sin is
the y-coordinate of the point A and cos is its x-coordinate. For angle , it is easy
to compare these definitions with the ones above; namely, the sine of an angle is the
length of the side opposite that angle divided by the hypotenuse. The side opposite
angle is the y-coordinate of point A, and the hypotenuse is 1, since A is on a unit
circle that is defined to have radius 1.
In general, if we have an angle like centered at the origin and whose first edge
goes through I = (1, 0) and whose other edge goes through point A = (x, y), then we
can define the trigonometric functions using Table A.4.
sine:
cosine:
tangent:

sin = y
cos = x
tan = y/x

cosecant: csc = 1/y = 1/ sin

secant:
sec = 1/x = 1/ cos
cotangent: cot = x/y = 1/ tan

Table A.4: Alternative Definitions of the Trigonometric Functions

There are advantages to this definition over the previous one. First, it is much
easier to work with coordinates, and second, this definition is valid for all anglesnot
only those that are bigger than 90 , but for negative angles, and for angles larger than
360. Of course the trigonometric functions for an angle like 390 are not terribly
interesting since it is the same as going around the circle once completely and then 30
more, so every trigonometric function will have thesame value for 390 as for 30 :
sin 390 = sin 30 = 1/2; cos 390 = cos 30 = 3/2, sec( + 360) = sec , et
cetera.
To better visualize the trigonometric functions, load the file Review/TrigFuncs.T
and press the Run Script button. Each of the six basic trigonometric functions is represented by a segment of a different color, and as the script runs, the angle cycles around
the whole circle and the segments change length accordingly. This demonstration, in
particular, makes it obvious what is meant when we say that the tangent (or cotangent
or secant or cosecant) goes to infinity, why sine and cosine are always between 1
and 1, et cetera.
Notice that if the angles are between 90 and 180 (in other words, if the angle
lies in quadrant II), the x-coordinates are negative, so the cosines and tangents of those

APPENDIX A. MATHEMATICS REVIEW

330

angles will be negative. If the angle is between 180 and 270 (in quadrant III), both
coordinates are negative, making the sine and cosine negative (but the tangent positive,
since it is the quotient of two negative numbers).
There are a lot of other things you can see from the figure, such as sin =
sin(180 ). This is because if you measure an angle counterclockwise from (1, 0)
you come to a point that has the same y-coordinate as you get if you measure that same
angle clockwise from (1, 0) (which is how you get to the angle 180 ). Similarly,
the pythagorean theorem shows instantly that sin2 + cos2 = 1 since the sine and
cosine are the x- and y-coordinates of a point on a circle of radius 1.
Notice that although the sine and cosine always make sense, there are times when
the tangent, cotangent, secant, and cosecant do not; namely, when the denominators in
their respective definitions are zero. Sometimes we just write something like tan 90 =
to indicate that as the angle gets closer and closer to 90 , tan gets larger without
bound. Be careful, however. If gets close to 90 from below, tan does get larger
and larger, but if approaches 90 from above, then tan gets to be a larger and larger
negative number. The same sort of thing holds true for all the other basic trigonometric
functions that go to infinity.

A.10.2  Key Theorems from Trigonometry

A.10.2.1 Some Concrete Trigonometric Values
Here is a short table that lists some exact values of the trigonometric functions for a
few interesting angles. When a trigonometric function is undefined as described in the
final paragraph in the previous section, appears in the table, so be sure to remember
what that means.

0
30
45
60
90

sin
0
1/2

2/2
3/2
1

cos
1
3/2
2/2
1/2
0

tan
0
3/3
1
3

cot

3
1

3/3
0

sec
1
2 3/3
2
2

csc

2
2
2 3/3
1

 Sometimes it is useful to know a few other key values. To save space, only the
sine and cosine values are listed; you can calculate the other values easily from these.
Note that many other trigonometric values for integer angles can be computed from the
tables above and below and using the formulas in Section A.10.2.2. For example:
sin(3 ) = sin(18 15 ) = sin(18 ) cos(15 ) cos(18 ) sin(15 ).

A.10. TRIGONOMETRIC DEFINITIONS

15

18
36
54
72
75

sin

1
2)
4( 6

1
4 ( 5 1)
p

1
10 2 5
4

1
5)
4 (1 +
p

1
10 + 2 5
4

1
2)
4( 6 +

331
cos

1
( 6 + 2)
4p

1
10 + 2 5
4

1
5)
4 (1 +
p

1
10 2 5
4

1
4 ( 5 1)

1
4 ( 6 2)

A.10.2.2 Basic Trigonometric Identities

Without proof (and it is a great exercise to do the proofs), here is a list of equations
satisfied by various of the trigonometric functions:
If + = 90 (if and are complementary8):
sin = cos

tan = cot

sec = csc .

For supplementary angles we have:

sin = sin(180 )
cos = cos(180 )

tan = tan(180 )

csc = csc(180 )
sec = sec(180 )

cot = cot(180 )

These are direct consequences of the fact that the trigonometric functions are defined in terms of coordinates of points on the unit circle:
sin2 + cos2 = 1

sec2 = tan2 + 1

csc2 = cot2 + 1

Here are the angle addition formulas9:

sin( )

cos( )

tan( )

= sin cos cos sin

= cos cos sin sin

tan tan
=
1 tan tan

(A.5)
(A.6)
(A.7)

See the diagram Proofs/CosineSummation.T for a Geometer proof of the sum of

angles formula for the cosine.
8 Note the letters co in the words:
sinecosine, tangentcotangent, secantcosecant, and
complementary. Coincidence? Not likely.
9 The symbol is the opposite of . It means that if you choose the + option of , you must use the
option in and vice-versa. The use of and in equation A.6 is equivalent to:

cos( + )

cos( )

cos cos sin sin

cos cos + sin sin

APPENDIX A. MATHEMATICS REVIEW

332

From these, it is easy to derive the multiple angle formulas since sin(2) = sin(+
), sin(3) = sin(2 + ), et cetera10 :
sin(2)
cos(2)
tan(2)
sin(3)
cos(3)
tan(3)

= 2 sin cos
= cos2 sin2 = 2 cos2 1
2 tan
=
1 tan2
= 3 sin 4 sin3
3

= 4 cos 3 cos
3 tan tan3
=
1 3 tan2

(A.8)
(A.9)
(A.10)
(A.11)
(A.12)
(A.13)

From these equations, it is easy to derive the following useful half-angle formulas11 :
r
1 cos
(A.14)
sin(/2) =
2
r
1 + cos
cos(/2) =
(A.15)
2
sin
1 cos
tan(/2) =
=
(A.16)
1 + cos
sin
Here are a few equations that are occasionally very useful:
 + 
 
sin + sin = 2 sin
cos
2
2
 
 + 
sin sin = 2 cos
sin
2
2
 + 
 
cos + cos = 2 cos
cos
2
2
 + 
 
cos cos = 2 sin
sin
2
2
sin( + )
tan + tan =
cos cos
sin( )
tan tan =
cos cos
From those we can also derive the following related formulas:


sin cos =
sin( ) + sin( + ) /2


cos cos =
cos( ) + cos( + ) /2


sin sin =
cos( ) cos( + ) /2

(A.17)
(A.18)
(A.19)
(A.20)
(A.21)
(A.22)

(A.23)
(A.24)
(A.25)

10 The formulas for cos n are closely related to the Chebychev polynomials: If x = cos , T (x) =
n
cos n. T0 (x) = 1, T1 (x) = x, T2 (x) = 2x2 1, T3 (x) = 4x3 3x, et cetera. It is an interesting
exercise for a student to investigate these polynomials.
11 Depending on the size of in formulas A.14 and A.15, the positive or negative value of the square root
must be used.

A.10. TRIGONOMETRIC DEFINITIONS

333

A.10.2.3 Area of a Triangle

One trivial application of trigonometry to
geometry is a simple formula for the area
of a triangle. If a triangle has two sides of
lengths a and b, and the included angle is
, then the area of the triangle is given by
(ab sin )/2.

B
a

Figure A.34 shows why. Drop a perpendicular line from vertex B to line AC,
Figure A.34: Area of a Triangle
intersecting AB at point D. By simple
Review/Triarea.T [P]
trigonometry, the length of BD, the altitude
of the triangle, is a sin . The area of the
triangle is half the base times the height, or (ab sin )/2.
A.10.2.4 The Law of Sines
The law of sines is one of the most useful theorems that relates the trigonometric functions to arbitrary triangles.
Theorem A.3 (Law of Sines) In any triangle 4ABC,
CA
AB
BC
=
=
.
sin A
sin B
sin C

D
O
A

In figure A.35, construct the circle that

circumscribes the given 4ABC by finding
its circumcenter O. Next, construct the diameter AD through O, and connect point
D with point B. Since AD is the diameter
of the circle, ABD is a right angle, and
hence sin ADB = AB/AD. ADB =
ACB = C since both of them are inscribed in the same arc AB. Hence AD =
AB/ sin C. A Geometer proof of the law
of sines appears in Proofs/Sinethm.T.


Exactly the same argument can be made

about every other angle in the trianglethat
the side opposite that angle divided by the
Review/Sinethm.T [P]
sine of the angle is the diameter of the circumscribed circle. But that diameter is constant, so all three quotients must be equal,
and the law of sines is proved.
Figure A.35: Law of Sines

Actually the law of sines often includes one more equality as follows:
CA
AB
BC
=
=
= 2R,
sin A
sin B
sin C

APPENDIX A. MATHEMATICS REVIEW

334

where R is the radius of the circumcircle for 4ABC. This is easy to see using the
same figure. ACB = ADB since they both subtend the same arc, but 4ABD is a
right triangle, so ABD = 90 . Thus
AB
AD
AD
AB
=
=
=
= AD = 2R.

sin C
sin D
sin 90
1
A.10.2.5 Angle Bisectors in Triangles
The law of sines gives a nice, useful result about angle bisectors in a triangle. In
figure A.36, we see 4ABC, where the
line AD bisects BAC. Let DAB =
CAD = and ADB = , so we
know that ADC = 180 .

Two applications of the law of sines

give:
sin
DB
sin
DC

=
=

sin
AB
sin(180 )
sin
=
.
AC
AC

180-
C

Figure A.36: Angle Bisector Property

Review/Anglebisect1.T [P]

Since sin = sin(180 ), we get:

sin
DC
DB
=
=
.
AB
sin
AC
There is another easy proof of this result that does not use trigonometry. Extend
the line AD and find point E on the extension such that CD = CE. Then 4CDE is
isosceles and from that it is easy to show that 4ADB 4AEC.
A.10.2.6 The Law of Cosines
Theorem A.4 (Law of Cosines) In any 4ABC, let a, b and c represent the lengths of
the sides opposite angles A, B and C, respectively, as in figure A.37. (In the figure, the
side of length b is broken into two parts having lengths l and b l.) Then
c2 = a2 + b2 2ab cos C.
Although the law of cosines is true in general, let us just prove it for the special
case where C is acute. A tiny modification of the proof will show that the theorem
is also true for obtuse angles. If C is a right angle, then cos 90 = 0, so the law of
cosines reduces to the pythagorean theorem: c2 = a2 + b2 .

A.10. TRIGONOMETRIC DEFINITIONS

As shown in figure A.37, drop a perpendicular from B to the line AC meeting it at
point D. Let D divide the side AC into two
segments of lengths l and b l. Denote the
length of BD by h.
Using the pythagorean theorem twice on
triangles 4ABD and 4BDC, we obtain:

335

B
c

h
A

b-l

Figure A.37: Law of Cosines

Review/Costhm.D [D]

c2 = h2 + (b l)2
and
a2 = h2 + l 2 .

If we eliminate h2 from these two equations, we get:

c2 = a2 l2 + b2 2bl + l2 = a2 + b2 2bl.

But l = a cos C, so we have:

c2 = a2 + b2 2ab cos C,
completing the proof for C an acute angle. Be sure to complete the proof for the case
where C is obtuse.
The law of cosines looks quite similar to the pythagorean theorem, and in fact,
except for the term 2ab cos C, it is identical.
Here are some slightly oddball formulas that, like the laws of sines and cosines,
relate the lengths of the sides of a triangle with some trigonometric functions of the
angles12 . Remember that the equations below apply to any sides or angles in a triangle

12 Since they are unusual, if you are teaching a trigonometry class, these may provide some examination
or extra credit questions.

336

APPENDIX A. MATHEMATICS REVIEW

so each of the equations below could be rewritten in two additional forms.

a = b cos C + c cos B
tan((A + B)/2)
a+b
=
ab
tan((A B)/2)
cos((A B)/2)
a+b
=
c
sin(C/2)
sin((A B)/2)
ab
=
c
cos(C/2)
r
(s b)(s c)
sin(A/2) =
bc
r
s(s a)
cos(A/2) =
bc
s
(s b)(s c)
tan(A/2) =
s(s a)
a b
cot (C/2).
tan((A B)/2) =
a+b

(A.26)
(A.27)
(A.28)
(A.29)
(A.30)
(A.31)
(A.32)
(A.33)

A.11 Coordinate Geometry

Applying coordinates to points and then simply using algebraic manipulation can be
a powerful technique to solve problems that seem totally geometric. The calculations
often become quite unwieldy, however. Today that is less of a problem since there
are good, commercially available computer algebra programs such as Maple and
Mathematica. A description of the use of such programs is beyond the scope of
this book.
Coordinate geometry (often called analytic
geometry) provides a method to convert state7
ments about geometry into statements about al6
B
gebra, and then to do algebraic manipulations to
5
4
arrive at results.
3
A
P
The idea is simplejust draw the geometric
2
figure in a standard Cartesian coordinate system
1
where algebraic variables are used for variable
1 2 3 4 5 6 7
points. Then any standard calculations from algebra are allowed.
Figure A.38: Distance: Point to
Coordinate geometry is based on the pythaPoint
gorean theorem that allows you to calculate the
Review/Coord1.D [D]
distance between a pair of points if you are given
their coordinates.
In figure A.38 coordinates have been assigned to the plane, and we can see that the
point A has coordinates (2, 2) and those of B are (6, 5). What is the distance between

A.11. COORDINATE GEOMETRY

337

points A and B? The segment AB is the hypotenuse of a right triangle whose right
angle is at the point P = (6, 2). The horizontal distance from P to A is 6 2 = 4.
Similarly, the vertical distance from P to B is 5 2 = 3. In the
right triangle 4ABP ,
we have P A2 + P B 2 = AB 2 , or 32 + 42 = AB 2 , or AB = 32 + 42 = 5.
In general, if the coordinates of any
two points A and B are (xA , yA ) and
(xB , yBp), then the length of the segment
AB is (xA xB )2 + (yA yB )2 .

C=(c,d)

(c/2, d/2)
((b+c)/2, d/2)
M
There is a fair amount of machinery
that can be used to do coordinate geometry
A=(0,0)
(b/2, 0)
B = (b, 0)
proofs. To do coordinate geometry effectively you need to know the equations of
lines, of circles, how to calculate intersec- Figure A.39: Analytic Proof of Three
tions of lines and circles, the conditions on Medians Theorem
the slopes of lines for them to be perpenReview/Coord2.D [D]
dicular, et cetera. All of these topics are
covered in a standard high school algebra
course so without derivation, a list of the most useful formulas can be found in Table A.5.
As an illustration of the technique, we will assume that you remember all of those
formulas, and we will prove that the three medians of a triangle meet at a point that is
2/3 of the distance from any vertex to the opposite side.
A nice thing about using coordinate geometry is that you can choose a coordinate
system that is convenient. To prove the theorem, there is no need to orient the triangle
in a totally arbitrary way (although the proof could still be done). It is easiest if we
twist around our arbitrary triangle until one of its vertices is at the origin and the other
is on the x-axis. In figure A.39 we have done just that. Point A is at the origin, B is at
(b, 0), and C is at (c, d).
The midpoints of the sides are easy to calculatetheir coordinates are the averages
of the coordinates of the two endpoints of the segments, so they are as shown in the
figure: (b/2, 0), ((b + c)/2, d/2), and (c/2, d/2).
The equations of the median lines originating at A, B, and C, respectively are:

d
x
b+c
d
(x b)
c 2b
2d
(x b/2)
2c b

(A.34)
(A.35)
(A.36)

APPENDIX A. MATHEMATICS REVIEW

338

(x2 x1 )2 + (y2 y1 )2
x + x y + y 
1
2
1
2
,
2
2

Distance between the points (x1 , y1 ) and

(x2 , y2 ).
Coordinates of the midpoint between the
points (x1 , y1 ) and (x2 , y2 ).

|ax1 + by1 |

a2 + b 2 + c 2

Distance between the point (x1 , y1 ) and

the line whose equation is ax+by+c = 0.

ax + by + c = 0

Completely general equation for a line (including vertical lines). Either a or b must
be non-zero.

y = mx + b

Equation of a line that intersects the y-axis

at b and has slope m.

y y1 = m(x x1 )

Equation of a line passing through the

point (x1 , y1 ) and having slope m.

y y1 =

y y 
2
1
(x x1 )
x2 x 1

y y1 =

 1 
m

(x x1 )

Equation of a line passing through the

points (x1 , y1 ) and (x2 , y2 ).
Equation of a line passing through the
point (x1 , y1 ) and perpendicular to a line
having slope m.

(x x1 )2 + (y y1 )2 = r2

Equation of a circle with center (x1 , y1 )

and radius r.

ax2 +bxy+cy 2 +dx+ey+f = 0

Completely general equation of a conic

section.

Table A.5: Useful formulas for coordinate geometry

A.11. COORDINATE GEOMETRY

339

Solving equations A.34 and A.35 simultaneously:

d
d
x =
(x b)
b+c
c 2b
1
1
x =
(x b)
b+c
c 2b
(c 2b)x = (b + c)(x b)
(c 2b b c)x = b(b + c)
3bx = b(b + c)
b+c
x =
3

plug this result into equation A.34:

 d  b + c  d
y=
= .
b+c
3
3

If we solve another pair of equations, say equations A.35 and A.36, we find that we
get exactly the same result, so the coordinates of the intersections of any pair of lines
is the same, and we can give a single name M to that point. M = ((b + c)/3, d/3).
Finally, to show that M is 2/3 of the distance between any vertex of the triangle
and its opposite side, use the pythagorean theorem to write p
down the distances, and
((b + c)/3)2 + (d/3)2
then use algebra to show that the distances
are
equal.
M
A
=
p
2
2
and the length of the median from A is ((b + c)/2) + (d/2) . We need to show that
M A is 2/3 the length of the median, or that
p
p
((b + c)/3)2 + (d/3)2 = (2/3) ((b + c)/2)2 + (d/2)2 .
Square both sides and expand:

(b2 + 2bc + c2 + d2 )/9 = (4/9)(b2 + 2bc + c2 + d2 )/4,

which is clearly true.
Do the same calculation for the other medians, and that will complete the proof.
This seems quite ugly, but at least it is almost completely mechanical. Geometric
calculations done by computers are done in exactly this way, and in fact whenever you
are dragging around points in Geometer, hundreds of equations very similar to those
above are being solved by the computer over and over.

A.11.1 Analytic Area of a Polygon

Theorem A.5 (Area of a Polygon) If the planar coordinates of the vertices of an nsided simple polygon are given by (x0 , y0 ), (x1 , y1 ), . . . , (xn1 , yn1 ), then if we
agree to define (xn , yn ) = (x0 , y0 ), the area of the polygon is given by the formula:
n1

1 X
A=
xi yi+1 xi+1 yi .
2 i=0

APPENDIX A. MATHEMATICS REVIEW

340

This formula gives a signed area in the sense that the value will be positive if the
points trace the polygon in a counterclockwise order and negative otherwise.
The formula holds for non-convex polygons, but does not hold if the edges of the
polygon cross each other.
We will complete the proof in a series of steps.
Lemma A.1 (Area of a Triangle with one vertex at the origin) If a triangle has vertices with coordinates (0, 0), (x0 , y0 ), and (x1 , y1 ) then its (signed) area is given by
(x0 y1 x1 y0 )/2.

B = (x1, y1)
A = (x0, y0)

Figure A.40: Area of Triangle at Origin

Review/Triorigin.D [D]

See Figure A.40. We would like to find the area of 4OAB where A has coordinates
(x0 , y0 ) and B has coordinates (x1 , y1 ). We know that
A(4OAB) = A(4OCB) + A(CDAB) A(4ODA).
It is easy to write down the areas of the two triangles and the trapezoid CDAB on
the right side of the equation:
A(4OAB)

=
=
=

(x0 x1 )(y0 + y1 ) (x0 y0 )

x1 y 1
+

2
2
2
x0 y 0 + x 0 y 1 x 1 y 0 x 1 y 1
x0 y 0
x1 y 1
+

2
2
2
x0 y 1 x 1 y 0
.
2

Notice that this formula is a signed area. Here we visit the points O, A, and B
in a counterclockwise order and obtain a positive area. For a clockwise ordering, the
area is negative since the area of the trapezoid will have the term (x1 x0 ) instead of
(x0 x1 ), and the triangle whose area was added is subtracted and vice-versa.

With this formula for the (signed) area of a triangle, we can proceed to a proof of
the final theorem.

A.11. COORDINATE GEOMETRY

341

P1

P2

P4

P0

B
P3
P5

O
Figure A.41: Area of a General Polygon
Review/Polyarea.D [D]

We will not give a completely rigorous proof, but it should be obvious what such
a proof would look like from our examination of Figure A.41 which shows a polygon
with 6 vertices: P0 , P1 , . . . , P5 . Assume the coordinates are given by Pi = (xi , yi ).
Draw a segment from each vertex Pi to the origin. From the lemma, we have a
formula for the area of 4OP0 P1 , for 4OP1 P2 , and so on, up to 4OP5 P0 . Adding
together the formulas for the individual polygons, and keeping in mind the signed
areas, it is easy to see that after all the additions and subtractions, the net result will be
whatever is contained within the polygon P0 P1 P2 P3 P4 P5 .
In this example, the areas of the triangles 4OP0 P1 , 4OP1 P2 , 4OP3 P4 , and
4OP5 P0 will be positive, and the areas of the other two will be negative.
The maze of lines heading back to the origin break up the plane into a bunch of
regions; let us just look at a couple of them, labeled A and B in the figure. Region
A should be inside the polygon. Its area is counted twice positively since it is inside
4OP3 P4 and 4OP5 P0 , but its area is subtracted once since it is inside the clockwiseoriented 4OP4 P5 . Thus its area is counted once, as it should be.

Region Bs area is counted once positively since it is inside 4OP5 P0 and once
negatively since it is inside 4OP4 P5 . Its net contribution is zero which is correct since
it is outside the polygon. Check some of the other regions to see that the same counting
scheme works.
To see why it works in general, imagine one of these regions in the plane. We
would like to count how many times its area is included in the grand sum, and see how
the positive and negative contributions of various triangles are combined.
Suppose the polygon does not include the origin, as in the figure. Imagine starting
at some particular point in the region your are testing and follow a straight line back to
the origin. Every time you cross a line it means that you have entered another triangle,
and if you were inside the polygon before crossing, you will be outside afterwards and
vice-versa. Each time you cross a line that enters the polygon, you must have crossed
a line putting you inside a counterclockwise oriented triangle and every time you leave

APPENDIX A. MATHEMATICS REVIEW

342

the polygon, you enter a clockwise oriented polygon. Thus, there is an alternation of
adding and subtracting until you get to the origin that you know is outside.
When you get to the origin, if you were initially outside, you will have crossed an
even number of lines, so the alternation will guarantee that the grand sum of areas for
your point will be zero; otherwise, the grand sum will be onethe area your point was
within was counted exactly once.
If the origin happens to be inside the polygon, the only change to the argument is
that an even number of crossings implies that you were initially inside.
Now, if we just add up the areas (remembering that P6 = P0 ), we obtain:

1 X
xi yi+1 xi+1 yi ,
A(P0 P1 P2 P3 P4 P5 ) =
2 i=0
5

and of course there is nothing special about the 6 vertex case.

A.12 Vectors
Vectors are usually covered in high school mathematics courses, but only at a very
superficial levelstudents generally have no opportunity to practice solving problems
with them. For this reason, the coverage of vectors here will be slightly different than
a review. It will consist of a review followed by a few problems solved using vector
techniques.
Vectors provide an intermediate method for solving geometric problems that is
partly symbolic and partly geometric. Problems approached with vector calculations
are often easy to solve and check, but they can also be be translated back to a geometric picture. Vectors can be defined in any number of dimensions, but we will deal here
with those in the two-dimensional plane.
There are two types of vectors, bound and unbound. Both indicate a length
and direction, but the bound sort also include a starting point. If we consider all the
bound vectors with the same direction and length as being the same, the class of all
such vectors is like a single unbound vector. Both types have their uses. If you simply
need to show that two segments have the same length and are parallel, you need to
show that they are the same unbound vector. If you would like to show that a point is
at the midpoint of two others, it is easier to use bound vectors.
The same distinction is important in physical problems. Unbound vectors are used
to describe relative movements, but if a vector represents a force, it is very different to
apply a force to the center of a rod or to its tip. Force on the center will drag the rod;
force on the tip will make it spin.

A.12.1 Unbound Vectors as Movements

A nice way to think of an unbound vector is as a description of a linear movement. A
complete description requires only a direction and a distance. The starting position is

A.12. VECTORS

343

unimportant, so go 5 kilometers northeast is basically a vector. This instruction is

independent of where you start, so could be applied to an object anywhere on the plane.

B
F

A
E
Figure A.42: Equivalent Vectors
Review/Vectors.D [D]

If vectors are movement descriptions, then it is easy to represent them as arrows.

The direction of the arrow indicates the direction of motion and the length of the arrow indicates the distance to move. If A and B are two points in the plane (as in
figure A.42), then the arrow with its tail at A and tip at B represents the vector
whose length is the length of the segment AB and whose direction is from A toward

B. We would write that particular vector as AB. In figure A.42, AB, CD and EF all
represent exactly the same vectorthey go in the same direction for the same distance.
The starting point is unimportant.
Equivalently, if an arrow representing a vector is moved parallel to itself anywhere
in the plane, it still represents the same vector.
In this book, we will always represent a vector with an arrow drawn over it. If we

happen to know the points at the tail and tip, we will write it as AB where A is the
tail and B is the tip. If we are just talking about a vector whose tip and tail may be

unknown, we can still give it a name like V . In some sense, the AB notation is a bit
misleading since in reality it stands for that particular length and direction anywhere
on the planenot just from the particular point A to the particular point B.
If vectors are instructions for movement, then adding two vectors is the same as
adding the instructionsjust do the first movement followed by the second. If one
vector is go north 1 kilometer and a second vector is go west 1 kilometer, then
to add them, first go north 1 kilometer and thenturn west and go one more kilometer.
The net result is a new vector: go northwest 2 (= 1.4142135 . . .) kilometers. It
should be clear that the combined instructions also indicate a distance and a direction,
still independent of the starting point.

Adding vectors using the arrow representation is easy: to add vector V1 to vector

V2 ,

slide V2 parallel to itself until its tail coincides with the tip of V1 . An arrow that

connects the tail of V1 to the tip of V2 represents the sum V1 + V2 . This is illustrated
in figure A.43.
When the vectors are placed tip to tail like this, it is obvious why the addition works
since this method clearly combines the two commands for motion.
One possible movement description is do not move. This is also a vector, the

zero vector, sometimes written as 0 . The zero vector is not the same as the number

APPENDIX A. MATHEMATICS REVIEW

344

V2
V1+V2

V2

V1
Figure A.43: Adding Vectors
Review/Vectors1.D [M]

zeroit is a vector representing a command not to move. It is clear that for any vector

V , we have: V + 0 = 0 + V = V . Doing something then doing nothing is
equivalent to just doing the something.
The negative of a vector is a vector going in the opposite direction for the same

length. In other words, AB= BA. This makes sense, since negation generally
means do the opposite, and the opposite of going from A to B is going from B to
A. As usual, subtracting vectors is the same as adding the negative of a vector, so to

calculate V1 V2 we add the negative of V2 to V1 to obtain V1 +( V2 ). Clearly

every vector V satisfies V V = 0 .

It does not matter in what order you add vectors: V1 + V2 = V2 + V1 . In other

words, it does not matter if you go north a kilometer and then east a kilometer, or if
you first go east a kilometer and then north a kilometer.

Vectors can be multiplied (or divided) by real numbers: 3 AB is a vector in the same

direction as AB, but 3 times as long. Similarly, AB /2 is again in the same direction,
but only half as long. Multiplying by negative numbers reverses the direction as well,

so 2 AB is a vector twice as long as AB and going in the opposite direction.

Since for most people, it is not obvious, at first, how to apply these vector methods
to geometric problems, we will begin with three examples worked in complete detail.
Read through these, and then try some of the additional examples at the end of the
chapter.

A.12.2 A Simple Vector Example

As a first example, we will use vectors to show that in any triangle 4ABC, the segment
connecting the midpoints of AC and CB is parallel to and half the length of segment
AB. See figure A.44.
Let M and N be the midpoints of AC and CB, respectively. Then in vector notation:

AB=AM + M N + N B .

A.12. VECTORS

345

We know that AM =AC /2 and N B=CB /2, so:

(A.37)

AB=AC /2+ M N + CB /2.

But we also know that AB=AC + CB so if

we divide all the vectors by 2, we obtain:

C
N

(A.38)

AB /2 =AC /2+ CB /2.

Combining equations A.37 and A.38 yields:

AB=AB /2+ M N ,

A
Figure A.44: Triangle Midpoints
Review/Trianglemids.D [M]

or

AB /2 =M N .
This proves the statement. Because of the vector equality, not only is the length of
M N half of AB, but because they are vectors the directions are the same as well, so
M N and AB are parallel.

A.12.3 A More Interesting Vector Example

F
B

E
C

F
B

E
C

Figure A.45: Triangle Squares

Review/Trisquares.D [M]

Given 4ABC, construct a square on the outside of each side as in the drawing on
the left in figure A.45. If the centers of those squares are D, E, and F as in the figure,
prove that the segment F C is perpendicular to and the same length as segment ED.
On the right side of the same figure is a diagram with the extraneous lines removed.
It contains isosceles right triangles erected on all the edges of 4ABC.

APPENDIX A. MATHEMATICS REVIEW

346

Here is a proof, using vectors. We begin with a few obvious identities:

ED

ED

FC

FC

(A.39)

= EC + CD

(A.40)

= EA + AF + F B + BD

(A.41)

= F B + BD + DC

(A.42)

= F A + AE + EC

There are clearly many routes to choose from E to D or from F to C, but the
choices above all use lengths that are known to be equal to other lengths, due to the
fact that the outer triangles are isosceles.
Adding equations A.39 to A.40 and A.41 to A.42, we obtain:

2 ED

= EC + CD + EA + AF + F B + BD

2F C

= F B + BD + DC + F A + AE + EC

(A.43)
(A.44)

What we would like to show is that the vector ED, when rotated 90 in a clockwise

direction, is the same as vector F C.

We can rotate the vector 2 ED in equation A.43 by rotating the six vectors on the

right each by 90 and adding the results. Let ( V ) represent the result of rotating the

vector V by 90 in a clockwise direction. Since the lengths F A = F B, DB = DC

and EC = CA, it is easy to rotate every term. We just have to be careful to get

the directions right: (EC) =AE, (CD) =BD, et cetera. Continuing like this, we
obtain:

(2 ED) =AE + BD + EC + F B + F A + DC .

This is exactly the same as the formula for 2 F C in equation A.44 (with the vectors
rearranged), so we are done.

A.12.4 Three Equilateral Triangles

In figure A.46, let 4OAA0 , 4OBB 0 and 4OCC 0 be three equilateral triangles all
described in a clockwise orientation. Let the points K, L and M be the midpoints,
respectively, of segments A0 B, B 0 C and C 0 A. Show that 4KLM is equilateral.
We can use the same general approach that we followed in Section A.12.3. We

notice that if we can show that, say, LM when rotated 60 clockwise becomes LK we
are done, since both with have the same length, and they meet in a 60 angle.

Here are some expressions for LM and LK. These particular paths were chosen
because they all involve only parts of the known equilateral triangles. It is certainly

A.12. VECTORS

347

true that LM=LK + KM, but we do not know any of these lengths.

= LC + CC 0 + C 0 M

LM

LM

(A.45)

(A.46)

= LB + B O + OA + AM

LK

(A.47)

= LB + B B + BK

LK

(A.48)

= LC + CO + OA + A K

Adding equation A.45 to A.46 and A.47 to

A.48, we obtain:

2 LM

2 LK

(A.49)

+ B O + OA + AM

LC + CC 0 + C 0 M + LB 0

LB + B B + BK + LC

+ CO + OA0 + A0 K

(A.50)

A
A

Since L and K are the midpoints of the seg

ments B 0 C and A0 B, LC= LB 0 , C 0 M=

Figure A.46: Three Triangles

Review/Threetris.D [M]

AM and BK= A0 K. These equalities simplify equations A.49 and A.50 to:

2 LM

2 LK

= CC 0 + B 0 O + OA

(A.51)
(A.52)

= B B + CO + OA

Let ( V ) represent the vector obtained when V is rotated 60 in the clockwise

direction, so from equation A.51 we obtain:

(2 LM) = (CC 0 ) + (B 0 O) + (OA).

(A.53)

It is easy to see that (CC 0 ) =CO, (B 0 O) =B 0 B and (OA) =OA0 . Substitut

ing these into equation A.53 we can see that (2 LM ) = 2 LK and we are done.

A.12.5 Touching Triangles

The following example makes use of vectors in the usual way, but uses the fact that the
dot product of a pair of vectors is zero if and only if those two vectors are perpendicular.
The problem is this: Suppose two triangles, 4XOZ and 4P OQ share a point O.
Suppose further that OZ = OQ and OX = OP and that OQ OZ and OX OP
as illustrated in figure A.47. If T is the midpoint of XZ, show that OT P Q.

APPENDIX A. MATHEMATICS REVIEW

348

We will show that OT P Q by showing

that the vector dot product: OT P Q = 0.

We know that P Q = P O + OQ, that OT

= (OX + OZ)/2, and that OQ OZ =

0 = OP OX since they are perpendicular

pairs.

product OT P Q = 0, we can ignore the

factor of 2 in the expansion of OT in the paragraph above, or equivalently, we can show

S
P

Since we just want to show that the dot

Figure A.47: Touching Triangles

Review/ShowPerp.T [M]

that 2 OT P Q = 0.

P Q 2 OT = (P O + OQ) (OX + OZ).

When we multiply ot the expression on the right above, we obtain:

P O OX + P O OZ + OQ OX + OQ OZ,

the first and last terms of which are zero.

It is also clear from the diagram that P OZ = XOQ since each is a right
angle added to P OQ, so the two other dot products have equal values, given that the
vectors have equal lengths and make equal angles with each other. The signs of those

dot products, however, are opposite since P O = OP . This proves that OT P Q.

A.12.6 Bound Vectors and Coordinate Geometry

In this book we will consider all of our bound vectors to have their tail at the same
place. If we want to use coordinate geometry, that same place will be the origin of our
coordinate system: (0, 0).
If we wish to describe our bound vectors with coordinates and all of them have
their tail at the origin, then each vector is completely determined by the coordinates of
the tip. The point (x, y) thus stands for the bound vector whose tail is at the origin and
whose tip is at the point (x, y).

If the vectors V1 and V2 are represented by the points (x1 , y1 ) and (x2 , y2 ), respec

tively, the sum or difference: V1 V2 is represented by the point (x1 x2 , y1 y2 ).

Similarly, if is a real number, then the vector V1 is represented by (x1 , y1 ).

In particular, V1 is represented by (x1 , y1 ). Finally, (0, 0) represents the zero

vector.
Following is a list of some properties of bound vectors. Some are important and
will be used later; some are not particularly important in the future, but shed some

A.12. VECTORS

349

light on how vectors work. Sometimes the arrow notation is used and sometimes the
coordinate notation, depending on which makes the property clearest.

The average of two vectors ( A + B )/2 is a vector that goes from the origin
to a point exactly half way between the tips of the other two vectors. Thus if a

parallelogram has a vertex at the origin and two others at A and B , then the

vector A + B represents the fourth vertex, and ( A + B )/2 is the the center
of the parallelogram.

If is a real number and A and B are two vectors, then vectors of the form

(1 ) A + B represent all the vectors whose tips lie on the line connecting

the tips of A and B . If 0 1 then the points lie between the tips of

A and B ; otherwise, they lie outside. In fact, represents the fraction of the

distance the tip of the new vector is between A and B . If = 1/3, then

(1 ) A + B = (2/3) A +(1/3) B is (1/3) of the way from A

to B 13 . In the previous item, the average of the two vectors corresponds to

= 1/2half way between the two vectors.

In 4ABC, if vectors A , B and C represent the three vertices of the tri

angle, then ( A + B + C )/3 represents the centroid of the triangle. See

Section A.12.7.
p

The length of the vector A = (x, y) is given by x2 + y 2 . (This is just the

pythagorean theorem applied to the triangle that includes the origin (0, 0), the
point (x, 0), and the point (x, y). This right triangle has legs of lengths x and
y, and the hypotenuse is the length of the vector.)
p An unbound vector described
with tail (x1 , y1 ) and head (x2 , y2 ) has length (x2 x1 )2 + (y2 y1 )2 .

If (x, y) is a vector, then (y, x) is another vector of the same length and perpendicular to (x, y). In fact, (y, x) is the vector (x, y) rotated counterclockwise
by 90 .
If you rotate the vector (x, y) counterclockwise by an angle , the resulting vector is (x cos y sin , x sin + y cos ). The previous item is a special case of
this, where = 90 .
Vectors can be defined in any number of dimensions. In one dimension, they
simply represent directions to the left and to the right; a negative vector goes left
and a positive vector goes right. Only a single number is needed to represent
a one-dimensional vector. In three dimensions, three numbers are needed, and
these represent the motion in the x, y, and z directions, where the z-axis is
perpendicular to both the x-axis and to the y-axis.

13 Do

not get confused by the fact that the numbers seem to be reversed. (2/3) A + (1/3) B is 2/3

like A and 1/3 like B . It is more like A than B , so it is closer to A than to B . If you get confused,
think of the extreme conditions where = 0 or = 1.

APPENDIX A. MATHEMATICS REVIEW

350

If A0 is the same as A rotated clockwise by an angle and B 0 is the same as

B rotated clockwise by the same angle , then A0 + B 0 is the same as A + B

rotated clockwise by .

A.12.7 Example: Vector form of the Centroid

Suppose we have a triangle 4ABC, where A , B and C are vectors from our origin
to the vertices of the triangle. What are the coordinates of the centroid? Notice that
in the proof that follows, we never need to make use of the actual coordinates of the
vectors. In fact, although we think of our vectors here as being two-dimensional, this
proof works in any number of dimensions.
The centroid is at the intersection of any two medians, so let us use the medians

originating at C and A .

The median from C connects it with the midpoint of the vector ( A + B )/2. We
know that all vectors along that line are described by:

A + B
,
(A.54)
C +(1 )
2
for some value of . Similarly, the values of the vectors along the other median are
given by:
!

B + C
A +(1 )
,
(A.55)
2
for some value of .
The intersection will occur when those two vectors are equal, or when:
!
!

B + C
A + B
C +(1 )
= A +(1 )
.
2
2
If we multiply everything out and move all the terms to the left side, and gather like
terms, we obtain:
!
!
!

1 2

1 2
+ B
+ C
=0 .
A
2
2
2
For this to be identically zero, we need all three coefficients to be zero, and if you
solve the equations (at first is looks like there are three equations and two unknowns,
but the equations are not independent), we find that the only solution is = 1/3 and
= 1/3. Putting these values back into either equation A.54 or A.55, we find that the

vector describing the centroid is ( A + B + C )/3.

A.13. COMPLEX NUMBERS

351

A.13  Complex Numbers

This section provides only a tiny introduction to the huge topic of complex numbers.
You can learn much more in virtually any book on complex analysis, but Needhams
book (see [Needham, 1997]) is a particularly geometrically oriented book as you can
tell from its title, Visual Complex Analysis.
The reason complex numbers are so useful for geometers is that they provide yet
another method to describe the locations of points on a plane.

A complex number has the form a + bi, where i is the imaginary 1 (in other
words i2 = 1), and a and b are arbitrary real numbers. Complex numbers are a lot
like two-dimensional vectors but instead of having an x- and a y-coordinate, you can
think of them having a real coordinate and an imaginary coordinate.
Thus they are commonly plotted on a plane with a real and an imaginary axis,
where the real axis is drawn in place of the x-axis and the imaginary axis in place of
the y-axis. The origin represents the point 0 + 0i, and the point a + bi would be plotted
a units to the right and b units above the origin (with the obvious reversals if a and/or b
is negative). We usually use letters like z or w (or the Greek letter zeta) to indicate
arbitrary complex variables that include both a real and an imaginary part.
All the usual operations exist for complex numbersaddition, subtraction, multiplication, division (except by zero). Here are formulas for the basic operations:
(a + bi) (c + di) = (a c) + (b d)i
(a + bi)(c + di) = (ac bd) + (ad + bc)i
 ac + bd   bc ad 
(a + bi)
+ 2
i
=
(c + di)
c2 + d 2
c + d2

To derive the equation for division of complex numbers above, simply multiply numerator and denominator of the expression on the left by c di.

In addition
to these, many of our favorite functions also make sense, such as: sin z,
cos z, ez , z, log z, et ceteraalthough
some care must be taken to define these prop
erly, especially functions like z and log z which are multiple-valued functions.
Except for the function ez (see Section A.13.1) we will not need any of these here.
If we think of points on the Euclidean plane as single complex numbers, many
important geometric operations have simple representations in terms of mathematical
operations on the complex numbers.
For example, if we simply add the number c + di to every point on the complex
plane, this has the effect of a translation of c units to the right and d units up (where c
and d are positivenegative values, of course, translate left and/or down). If we multiply our numbers by cos + i sin this has the effect of rotating them counterclockwise
about the origin by an angle . There are many similar operations.
Similarly, the midpoint of z and z 0 is given by (z + z 0 )/2. If z1 , z2 , and z3 are the
vertices of a triangle, the centroid is given by (z1 + z2 + z3 )/3. Multiplication by a
pure positive real number scales the points toward or away from the origin uniformly.

APPENDIX A. MATHEMATICS REVIEW

352

For example, multiplication by 2 makes everything twice as big, and multiplication

by 1/3 makes everything one third as big. Multiplication by 1 reflects everything
through the origin (or equivalently, rotates it by 180 ), and multiplication by i rotates
counterclockwise by 90 .
If z = a + bi is an arbitrary complex number, then we say that z = a bi is the
complex conjugate of z. Geometrically, it amounts to the reflection of z across the real
line (the line containing all the complex numbers whose imaginary parts are zero.) The
complex conjugate has the following properties that are not hard to show:
(z w)
zw
z/w

= z w

= zw

= z/w.

The complex conjugate allows us to express the distance between a point and the
origin (0 + 0i), and this distance is written |z|, is defined by:

|z| = z z,
and is called the absolute value of z.
The formula above works because for any complex z, z z is always real and nonnegative. To see this, let z = a + bi where a and b are real numbers. Then
z z = (a + bi)(a bi) = a2 + b2 ,
which is certainly always non-negative. In fact, if you think of a and b as coordinates
on the complex plane, then we have
p
|z| = |a + bi| = a2 + b2 ,
which is exactly the same distance that the pythagorean theorem would give us for the
distance from a point z to the origin.
The distance between two complex numbers z and w is given by |z w|.

A.13.1  Eulers Formula

If you are not afraid to work with complex numbers, there are some formulas which
can simplify many of the trigonometric calculations substantially.
If is any real number, then Eulers formula states that:
ei = cos + i sin .
We saw the expression cos + i sin in the previous sectionmultiplication by this
number is equivalent to a rotation by an angle .
If that is true, then we also have:
ei = cos() + i sin() = cos i sin .

A.13. COMPLEX NUMBERS

353

The two equations above can be added or subtracted and manipulated in various
other ways to yield:

cos =

ei + ei
2

sin =

ei ei
2i

tan =

ei ei
i(ei + ei )

We will not spend much time on it, but here is an example of how Eulers formula can be used to show one of the trigonometric equalities above. Let us derive
equation A.6the formula for cos( + ):
ei(+) + ei(+)
2
ei ei + ei ei
=
2
(cos + i sin )(cos + i sin ) + (cos i sin )(cos i sin )
=
2
(cos cos sin sin ) + i(cos sin + sin cos )
=
2
(cos cos sin sin ) i(cos sin + sin cos )
+
2
= cos cos sin sin .

cos( + ) =

But if you are a little more lazy, you can get the formulas for both sin( + ) and
cos( + ) with less work:
ei(+) = ei ei = cos( + ) + i sin( + )
= (cos + i sin )(cos + i sin )
= (cos cos sin sin ) + i(cos sin + sin cos ).
Setting the real part of the right side of the first line equal to the real part of the final
line and similarly for the imaginary parts, we get both formulas:
cos( + )
sin( + )

= cos cos sin sin

= cos sin + sin cos .

354

APPENDIX A. MATHEMATICS REVIEW

Appendix

Geometer Art
At the top of the first page of each chapter of this book is a figure produced by Geometer that is primarily artistic. This appendix contains a short description of each. The
Geometer files and any C-language files that were used to generate the figures can be
found in the Art subdirectory of the Geometer installation.
Many of the files are parameterized so by editing the files and changing a number
or two different versions of the artwork can be generated. Here are the descriptions:
Chapter 0: Preface

Geometer file: Art/Koch5.T

C file: Art/koch.c
The Koch snowflake curve is the limiting case of a sequence of shapes that begins
with an equilateral triangle. To obtain the next shape in the sequence from the
previous, erect an outward-pointing equilateral triangle on the middle third of
355

APPENDIX B. GEOMETER ART

356

each segment of the boundary. Thus the second shape is like a star of David, et
cetera. The illustration here is the fifth step in this sequence.
The curve is interesting because in its limiting form, it encloses a finite area, but
has a boundary whose length is infinite.
Chapter 1: Introduction

Geometer file: Art/Pretty1.T

This pattern is the result of a Geometer programming bug. A cosine function

was applied instead of a sine function in an attempt to generate the Maurer rose
that appears as the art at the beginning of Chapter 9.
Chapter 2: Computer-Assisted Geometry
Geometer file: Art/dragon.T
C file: Art/dragon.c

This is the dragon curve. As you click on the Next button, successive approximations to the curve appear. The end result is a space-filling curve in the sense
that in the limit, every point of the region is approached arbitrarily closely by the
curve. The limiting length of the curve is, of course, infinite.
Chapter 3: Mathematics Review
Geometer file: Art/Caustic.T

This illustration shows how caustic curves are generated in optics. Light rays
from a point source are internally reflected within a sphere, and the resulting
pattern of sparse and dense concentrations appears. Light through a wine glass
generates similar patterns on a tablecloth. Press the Run Script button to generate
the caustic.
Chapter 4: Computer-Aided Proof This image is generated by beginning with
an array of 9 circles arranged in three rows and columns, and inverting each of
the circles in the other 8. Then those circles are again inverted in the original 9
circles and the process is repeated for a few steps. Different initial arrangements
of circles and different depths can be achieved by modifying the C file and
rerunning it.
Chapter 5: More Useful Theorems
Geometer file: Art/fourier.T

Press Run Script to show a sequence of trogonometric Fourier series that approximate a square wave. Eight steps of the approximation are superimposed.
Chapter 6: Locus of Points

Geometer file: Art/golden.T

A geometric illustration of one property of the golden section is illustrated here.

Each time the Next command is pressed, a square is chopped
off the end of a

rectangle whose sides are in the golden ratio: = (1 + 5)/2 : 1. Each time a

357
square is chopped off, the resulting rectangle has exactly the same shape as the
previous.
Chapter 7: Triangle Centers

Geometer file: Art/hilbert.T

C file: Art/hilbert.c
Each press of the Next button displays another in a sequence of curves whose
limit is the Hilbert space-filling curve.

Chapter 8: Inversion in a Circle

Geometer file: Art/Maurer.T

This figure is called the Maurer rose and is generated with the Run Script button.
It is generated by connecting two points that move around an n-leaved rose with
a fixed angular offset between them. The Geometer diagram can be modified to
change both n and the angular displacement.
Chapter 9: Projective Geometry
Geometer file: Art/peano.T
C file: Art/peano.c

The Peano curve is another space-filling curve. The Geometer file is generated
by a C computer program. Press Next to view successive approximations to
the true Peano curve.
Chapter 10: Harmonic Point Sets
Geometer file: Art/Poly/Poly.T
C file: Art/Poly/poly.c

This is the projection of the {3, 3, 5} 4-dimensional polytope onto a two dimensional surface. The Geometer diagram is generated by a C program that can
be modified to change the direction of the projection.
Chapter 11: Geometric Presentations
Geometer file: Art/square.T
C file: Art/square.c

This is another space-filling curve whose approximations are quite interesting.

Chapter 12: Geometer Proofs

Geometer file: Art/Whirl/Whirl.T

C file: Art/Whirl/Whirl.c
This is a simple illustration created by drawing a regular polygon (in this case a
regular heptagon) repeatedly drawn with a slight increment in the rotation angle
and a slight decrement in size.

APPENDIX B. GEOMETER ART

358
Appendix A: Geometric Construction
Geometer file: Art/epicycloid.T

One definition of an epicycloid is as the envelope of the straight lines connecting points on the circumference of a circle with points n times as far along the
surface. Press the Run Script button to view an example. If you edit the Geometer code, you can change the value of n to generate different epicycloids. The
multiple n need not be an integer, but the results are a bit less interesting.
Appendix B: Geometer Art
Geometer file: Art/Liss.T

In this pattern a pair of points follow each other with a fixed angular difference as
they trace out a lissijous curve. Those points are connected with a line segment.
Both the frequencies of the curve and the angular difference can be changed by
editing the Geometer file.
Appendix C: Geometric Problem Solving Strategies
Geometer file: Art/Spirograph.T

This is similar to the Spirograph example in the text (see Section 2.15.3) except with different radii and numbers of teeth, and in addition, it is drawn with
connected line segments instead of as a series of dots.

Appendix

Geometric Problem Solving

Strategies
It is impossible to give a cookbook method that will take an arbitrary proposed geometric theorem and either prove it or show that it is false. That this cannot be done was
proved in 1931 by Kurt Godel ([Godel, 1931]). In spite of that, it is certainly possible
to learn and organize geometric problem solving techniques. The more problems you
work, the more you will see certain techniques used.
There are many excellent books on general problem-solving techniques that can be
applied to any area of mathematics, but most do not concentrate on geometric problems.
Many of the best were written by George Polya, such as [Polya, 1957], [Polya, 1990a],
[Polya, 1990b], and [Polya, 1962]. There are many others, but a good recent one is
[Zeitz, 1999]. Section C.13 lists some of these more general mathematical approaches
to problems.
359

360

APPENDIX C. GEOMETRIC PROBLEM SOLVING STRATEGIES

This appendix is organized a bit like a thesauruslook up a problem category to

obtain a list of possible approaches to problems in that category. Obviously no such
listing (either of categories or strategies) could ever be complete, so it may not contain
exactly what you need.
To use this appendix, first have a clear idea of what you are trying to do, like:
Show that points A, B and C all lie on a line. or Find the length of segment AB.
or Construct a circle tangent to two circles and a line. Then look through the list of
general categories below to see which ones might apply. Go to the referenced sections
and continue from there.
Congruence: Methods to prove that geometric figures are congruent. Section C.1.
Similarity: Methods to prove that geometric figures are similar. Section C.2.
Special Figures: Showing that figures are special in some way. Is the figure
a square? A rectangle? An equilateral triangle? An isosceles triangle? Section C.3.
Concurrence: Methods to prove facts about concurrenceare two points in fact
the same point? Do three points lie on a line? Do four points lie on a circle? Do
three lines go through the same point? Are two circles tangent? Section C.4.
Measures: Methods to measure things. How long is a line segment? What is the
measure of an angle? What is the area of a triangle? Section C.5.
Equality and Inequality: Methods to show line segments, angles, areas, et
cetera, are of equal or different lengths. If they are different, which one is larger?
How to show triangles are equilateral or isosceles, how to show that a quadrilateral is a parallelogram, rhombus, or trapezoid. Section C.6.
Ratios: Methods to work with ratios: how to find them, how to calculate with
them, what to use them for. Section C.7.
Inversion: Methods to use inversion to simplify a proof or construction.
When to Draw the Line: When and how to construct auxiliary lines or circles
in a geometric figure to complete a proof.
Construction Techniques: Making straightedge and compass constructions.
Relabeling How to work out the details once, and then be able to say, et cetera.
Each section contains short descriptions of techniques, possibly followed by references to sections where that particular technique was used or described. Techniques
that are advanced or somehow difficult to use are preceded by .

C.1. CONGRUENCE

361

C.1 Congruence
Technically1, congruence is an undefined term. In spite of that, one can think of congruent figures as having exactly the same size and shape, but they may be flipped
upside-down or moved around or rotated.
1. Two segments are congruent (have the same length).
a. They are sums or differences of congruent segments.
b. They are corresponding parts (where part means any linear partedges,
altitudes, perimeters, et cetera) of congruent triangles.
c. They are opposite sides of a parallelogram.
d. They are the two parts of a line bisected by another line or point. (For
example, the diagonals of a parallelogram bisect each other.)
e. They are opposite equal angles in a triangle.
f. They are radii or diameters of the same circle.
g. They are the two external tangents to a circle from a point outside, or they
are a pair of common external tangents (or a pair of internal tangents) of
two circles.
h. Any point on an angle bisector is equidistant from the two sides of the
angle.
i. Segments that subtend equal arcs are congruent.
2. Two angles are congruent.
a.
b.
c.
d.
e.
f.
g.

They are vertical angles.

They are cut off by a line that is a transversal of two parallel lines.
They are corresponding angles in congruent triangles.
They are sums or differences of equal angles.
They are corresponding angles in similar triangles.
They are opposite equal line segments in a triangle.
They are both interior angles and are inscribed in the same arc (or equal
arcs) in a circle, or they are both central angles and subtend equal arcs
in a circle. (This is especially common for (right) angles inscribed in a
semicircle.)
h. They are complementary (or supplementary) equal angles.
i. They are opposite angles in a parallelogram.

3. Two triangles are congruent.

a. SAS (side-angle-side)show that two sides and the included angle in both
triangles are congruent.
b. SSS (side-side-side)show that all three sides of the two triangles are congruent in pairs.
c. ASA (angle-side-angle) or AAS (angle-angle-side)show that one side is
congruent and that two of the corresponding angles are also congruent. Be
1 In a pure axiomatic description of geometry, there are certain undefined terms, and all we can technically
know about them is what we can derive from the axioms and logic. The relation congruent to is one of
these undefined terms.

APPENDIX C. GEOMETRIC PROBLEM SOLVING STRATEGIES

362

sure that the angles are in the corresponding positionsin other words, if
the side is between the two angles in one, is also between them in the other,
and if it is not between them, it must form part of the same angle in both
triangles. (See
d. Hypotenuse-legshow that they are both right triangles and that the two
hypotenuses and two corresponding legs are congruent. Note: This is a
special case of the SSA theorem that states that we can conclude that the
triangles are congruent only if the angle is 90 or greater.
4. Two arcs are equal.
a.
b.
c.
d.

They have equal central angles.

They have equal inscribed angles.
They are sums or differences of equal arcs.
They are cut from a circle between a pair of parallel lines.

C.2 Similarity
Two figures are similar if they are the same shape, although they may be different sizes.
1. Two triangles are similar.
a. AA (angle-angle)show that two angles in one are the same as two angles
in the other. Notice that since any two equal angles imply the similarity of
a pair of triangles, there is no need for rules like ASA or AAS for similarity.
b. A line parallel to the base of a triangle through the other two sides cuts off
a triangle similar to the original. (This is basically the same as AA.)
c. All three pairs of corresponding edges in the two triangles are parallel.
d. SAS (side-angle-side)show that two sides in a triangle are in the same
ratio with the corresponding sides in another, and that the angles included
between those sides are equal.
e. SSS (side-side-side)show that all three sides in one triangle are in the
same ratio with the corresponding sides in another.
f. If two chords AB and CD of a circle intersect at a point P then 4AP C
4DP B.

C.3 Special Figures

1. A triangle is equilateral.
a. Show all three sides are equal.
b. Show all three angles are equal.
c. Show two angles are equal to 60 .
2. A triangle is isosceles.
a. Show that two sides are equal.
b. Show that two angles are equal.

C.4. CONCURRENCE

363

3. A quadrilateral is a rectangle.
a. Show three angles are 90 .
b. Show that it is a parallelogram with one 90 angle.
c. Show that it is a parallelogram and that the diagonals are of equal length.
4. A quadrilateral is a square.
a. Show that it is a rectangle with adjacent equal sides.
b. Show that all sides are equal and one angle is 90 .
5. A quadrilateral is a trapezoid.
a. Show that two sides are parallel.
6. A quadrilateral is a rhombus.
a. Show that it is a parallelogram and that the diagonals are perpendicular.
b. Show that all four sides are equal.
7. A quadrilateral is a parallelogram.
a.
b.
c.
d.
e.
f.
g.

Show that both pairs of opposite sides are parallel.

Show that both pairs of opposite sides are equal.
Show that its vertices are the midpoints of any quadrilateral.
Show one pair of sides is parallel and the same length.
Show that the opposite angles are congruent.
Show that adjacent angles are supplementary.
Show that the diagonals bisect each other.

C.4 Concurrence
1. Two points are concurrent.
a. Show that they both divide a line segment in the same ratio.
b. Show that each lies on two non-parallel lines.
2. Three points lie on the same line.
a. Points A, B, and C lie on the same straight line (with B between A and C)
if AB and BC make the same angle with some third line. (In a sense they
share a point and would normally be parallel, so they are the same line.)
b.  Use Menelaus theorem (see Section 5.2).
3. Three lines pass through the same point.
a. Show that they pass through a known point. For example, if the three lines
can be shown to be altitudes, or angle bisectors, or medians, et cetera, of
some triangle, they are concurrent.
b. Show that two of the lines pass through the same point of the third.
c.  Use Cevas theorem (see Section 5.1).
d.  Use Brianchons theorem.
4. Three circles intersect at a point.

364

APPENDIX C. GEOMETRIC PROBLEM SOLVING STRATEGIES

a. Show that the point of intersection of two of them lies on the third, perhaps
by showing it is on an inscribed triangle or quadrilateral.
b. Find a point whose distance to each circle is the radius of that circle.

5. A point lies on a circle.

a. The point is the same distance from the center as another point known to
be on the circle.
b. The point makes an equal angle with an arc of the circle as another point
known to be on the circle.
c. The point lies on a vertex of a right triangle whose hypotenuse is the diameter of the circle.
6. A line and a circle are tangent.
a. Show the line is perpendicular to a radius and passes through the endpoint
of that radius.
7. Two circles are tangent.
a. Show that the distance between their centers is the sum of their radii.
8. Four points lie on the same circle. (Equivalently, show that the quadrilateral
whose vertices are those points can be inscribed in a circle.)
a. An isosceles trapezoid can be inscribed in a circle.
b. If the points can be divided into two right triangles that share a hypotenuse,
they lie on the same circle.
c. More generally, if the four points are A, B, C, and D, show that ACB =
ADB.
d. If either pair of opposite angles in a quadrilateral are supplementary, the
four points lie on a circle. In other words, if ABCD is a convex quadrilateral and ABC + CDA = 180 , then the four points lie on a circle.
e.  (Ptolemys theorem) If ABCD is a convex quadrilateral and AB CD +
BC DA = AC BD then the four points lie on a circle. See Section 5.5.

9. A quadrilateral circumscribes a circle.

a. If the four sides, in order, have lengths a, b, c, and d, the quadrilateral

circumscribes a circle if and only if a + c = b + d.
b. Show that each side of the quadrilateral is tangent to the circle.

C.5 Measures
Many problems require that you find a measure of a length, angle, area, or something
else in terms of some given measurements. This is a huge category of problems, but
there are some useful general approaches:
Draw a pictureeither by hand or with Geometer. Label the parts you know
and look at the parts you do not.

C.5. MEASURES

365

What relationships do you see in the figure. Look for things like right angles,
parallel lines, centers of circles, equal included angles, angles that sum to a triangle or to a straight line, et cetera.
1. To find the length of a segment.
a. Use the pythagorean theorem (stated and proved in Section A.6.5) on a
right triangle.
b. Use the fact that the segment is the sum or difference of other segments that
you can measure.
c. Use the law of cosines (see Section A.10.2.6).
d. Use the law of sines (see Section A.10.2.4).
2. An angle is a right angle. (Equivalently, show that the two lines making up
the angle are perpendicular.)
a. Show that the angle is inscribed in a semicircle.
b. Use the pythagorean theorem: If the angle is opposite a side of length c in
a triangle whose other sides are a and b, show a2 + b2 = c2 .
c. A radius (or diameter) that bisects a chord of a circle is perpendicular to
it. Similarly, a tangent to a circle is perpendicular to the radius at the point
of tangency.
d. The diagonals of a rhombus or kite are perpendicular.
e. The line connecting the centers of two intersecting circles is perpendicular
to the line connecting the points of intersection.
f. The two angle bisectors of two intersecting lines are perpendicular.
g. Two lines that make the same angle with perpendicular lines are themselves
perpendicular.
3. Lines are perpendicular. (This is equivalent to showing they form a right
angle. See above.)
4. Lines are parallel.
a. Show that they have no points in common.
b. Show that a transversal line makes equal angles with both the given lines.
(The most common special case of this is to show that they are both perpendicular to the same line.)
c. Show that the lines are opposite sides in a parallelogram.
d. Show that both are perpendicular to the same line.
e. Show both are parallel to the same line.
f. A line connecting the midpoints of two sides of a triangle is parallel to the
third side.
g. Two lines that make the same angle with parallel lines are themselves parallel.
5. To find the measure of an angle.
a. The three angles in a triangle add to 180 .
b. The n internal angles in an n-sided polygon add to (n 2) 180 degrees.
c. Complementary angles add to 90 .

APPENDIX C. GEOMETRIC PROBLEM SOLVING STRATEGIES

366
d.
e.
f.
g.

Supplementary angles add to 180 .

Angles that go completely around a point add to 360 .
Angles that meet to form a straight angle add to 180 .
The central angle of an arc is twice an angle inscribed angle inscribed on
that arc.
h. A tangent to a circle with a secant at its point of tangency form an angle of
half the measure of the arc it cuts.
i. The exterior angle of a triangle is equal to the sum of the two interior
angles at the other vertices.
j. Opposite angles in a cyclic quadrilateral add to 180 .
6. To find the area of a polygon.
a. The area of a triangle is half the base times the height. Remember that
there are three different base-height combinations and any of them works.
b. Use Herons Formula to find the area of a triangle given the lengths of the
sides (see Section 5.6).
c. The area of a parallelogram (or rectangle, in particular) is the base times
the height.
d. If a complex figure can be divided into smaller figures, its area is the sum
of the areas of the smaller figures.
e. The area of a triangle that has sides of lengths a and b and included angle
is (ab sin )/2.
f. Use Brahmaguptas Formula (see Section 5.6.3) to find the area of a cyclic
quadrilateral.

C.6 Equality and Inequality

1. One segment is larger than another.
a. Show that the segments are opposite unequal angles in a triangle. The
larger segment is opposite the larger angle.
b. The hypotenuse of a right triangle is longer than either of the legs.
c. The sum of any two legs of a triangle is longer than the third. The sum is
equal only for degenerate triangles. (This is called the triangle inequality.)
2. A line is an angle bisector.
a. Show that both the angles formed are equal.
b.  In 4ABC, if the line through A hits the segment BC at a point D, and
if AD : BD = AC : CD then the line AD bisects BAC.

C.7 Ratios
1. Equality of products or ratios of lengths.

C.8. INVERSION IN A CIRCLE

367

a. Usually when products of lengths are to be proved equal, or ratios are to be

proved equal, it is done by finding one or a series of similar triangles whose
ratios are then multiplied or divided together to get the desired product or
ratio.
b. Three parallel lines cut equal ratios of any transversal line.
c. If two chords AB and CD of a circle intersect at a point P then P AP B =
P C P D.
d. In a triangle, the angle bisector divides the opposite side in the same ratio
as the lengths of the adjacent sides.
2. Finding Ratios where to get ratios of lengths.
a. Similar triangles (or any similar polygons) have equal ratios of all the
corresponding linear parts. The parts may be sides, medians, altitudes,
perimeters, et cetera.
b. Fixed ratios. For example, the diameter of a circle is always
twice the

radius, or the altitude of an equilateral triangle is always 3/2 times the

length of the side.
c. Chords AB and CD of a circle that cross at P yield AP P B = CP P D.
Note that P need not lie inside the circle for this to be true.
d.  Menelaus theorem yields products of ratios that are equal to 1 (see Section 5.2).
e.  Cevas theorem yields products of ratios equal to 1 (see Section 5.1).
f. Points that are inverted in the same circle yield ratios of lengths (ee Section 8.8).

C.8  Inversion in a Circle

Inversion is so important that an entire chapter (Chapter 8)is devoted to it. The examples in this section will not make much sense unless you are familiar with the contents
of that chapter.
This powerful technique converts circles and lines to other circles and lines, but
which are converted to which depends on the location of the center of inversion, and
on the radius of the circle of inversion. The technique is almost always the same
choose a center of inversion and a radius (sometimes the radius is unimportant), invert
the lines, circles, and points in the problem, solve the problem on the inverted figures,
and then re-invert that solution to get a solution of the original problem.
For reference, here is a short list of the more important properties of inversion
through a circle K centered at O:
1. Points are inverted to points.
2. Circles and lines are inverted to circles and lines, but a circle may become a
line and a line may become a circle.
3. Only lines or circles passing through O are inverted to lines.
4. Angles between lines and/or circles are preserved by inversion.
5. Inversion sends O to the point at infinity and the point at infinity to point O.

368

APPENDIX C. GEOMETRIC PROBLEM SOLVING STRATEGIES

6. Points on K are inverted to themselves.

C.9 Algebraic Manipulation

If you are good at algebra, it is often not too hard to assign a coordinate system of some
kind to your figure and then to state the various relationships algebraically and using
those, to prove the desired result using solely algebraic manipulations.
There are many coordinate systems, and almost all are amenable to algebraic manipulation. But there are coordinate-free representations (vectors) that can also be manipulated algebraically.
Often, one can prove properties with a combination of geometric and algebraic
techniquespurely geometric theorems are used to obtain various ratios and then those
ratios are manipulated algebraically to obtain the desired results.

C.9.1 Calculating with Angles

One of the most common uses of algebra in geometric proofs is for manipulation of
formulas involving angles. It is easy to write down many angle equalities in most
figuresthe angles in a triangle add to 180 or in a quadrilateral to 360 , opposite
angles in a cyclic quadrilateral add to 180 , et cetera.

C.9.2 Calculating with Ratios

It is also easy to find ratios and quite often a bit of algebra is needed to combine the
various formulas in a manner that gives the desired result.

C.9.3 Working with Areas

Often the area of a figure can be calculated in different ways, and those values can be
set equal. For example, figures can be divided into triangles in many ways, and every
sum of those triangle areas must be the same. Here are some places where areas are
used in algebraic manipulations:

C.9.4 Primarily Algebraic Theorems

The following theorems are geometric, but their results are typically applied algebraically:
1. The pythagorean theorem relates the lengths of the three sides in a right
triangle. See Section A.6.5.
2. Stewarts theorem relates the length of a cevian (a line connecting a vertex
to the opposite side) of a triangle to the other lengths. See Section A.6.6

C.10. WHEN TO DRAW THE LINE

369

3. Herons formula gives the area of any triangle in terms of the lengths of its
sides. See Section 5.6.
4. Brahmaguptas formula allows you to calculate the area of a cyclic quadrilateral given the lengths of its sides. See Section 5.6.3.
5. Ptolemys theorem relates the lengths of the sides and diagonals of a cyclic
quadrilateral. See Section 5.5.
6. The law of sines relates the sines of the angles of a triangle with the lengths
of the sides opposite them and to the diameter of the circumcircle. See Section A.10.2.4.
7. The law of cosines is a generalization of the pythagorean theorem to arbitrary triangles. See Section A.10.2.6.
8. Various trigonometric formulas. See Section A.10

C.9.5 Using Trigonometry

See Section A.11 for a discussion of the technique and an example where we show that
the three medians of a triangle meet at a point.

C.9.6 Vector Techniques

Section A.12 is a nice introduction to this technique, together with a demonstration that
the centroid of a triangle is the average of the three vectors to its vertices. Here are a
couple of situations where vector techniques might be used:
1. If there is rotational symmetry in the problem or solution. For example, if
you wish to prove that a triangle is equilateral, show that any edge can be
rotated 60 to become another edge.
2. If an entire figure is rotated or translated and relationships are sought between the figure and the rotated or translated version.

C.10 When to Draw the Line

One of the hardest techniques for beginners to learn is when it might be useful to draw
an auxiliary line or circle to complete a proof. Here are some rules of thumb:
1. If you want to apply some theorem about triangles, but your figure contains
no triangles, it may be useful to draw a line that makes a triangle. Obviously,
the word triangle in the previous sentence can be replaced by any other
geometric figure. A very common case is the construction of a line that
produces a pair of similar triangles.
2. If you need to calculate lengths or relative lengths, it is often useful to use
theorems like the pythagorean theorem. To do so requires a right triangle,
so to create one, you often need to drop a perpendicular or do something
else to make one or more right triangles in your figure.

370

APPENDIX C. GEOMETRIC PROBLEM SOLVING STRATEGIES

3. If you need to know something about a sum of lengths or angles, try to find a
construction that puts the lines end-to-end, or the angles next to each other
so that addition simply amounts to looking at a combined length or angle.
4. If there is some apparent concurrency that is not obvious from the statement of the problem, it is often useful to make a construction that somehow
helps you prove that the concurrency is in fact true. For example, if in your
drawings (with Geometer or drawn by hand) it seems like three points lie
on a line, or that four points lie on a circle, try drawing that line through
two of the points or the circle through three of the points and see if you can
prove that the other point lies on the line or circle.
You can sometimes discover concurrencies using Geometer by drawing a
suspicious point in a smearing color and then manipulating the rest of the
figure to see if that point moves in a regular wayalong a line or around a
circle, for example.
5. Remember that there is a whole class of problems that can be solved by
inverting your figures through a circle. A discussion of when to use inversion
can be found in this chapter in Section C.8.
6. The list above does not cover all cases. Here is a list of other miscellaneous constructions that have been used in this book, together with a short
description of how and why each was used.
a. In Section A.10.2.3, a perpendicular line is dropped to convert an arbitrary
triangle into one where we can use the definition of sin on a right triangle.
b. In Section A.10.2.4, we construct a circle and a right triangle inscribed in
it to be able to use our definition of sin on a right triangle.
c. In Section 12.1.3, a series of reflections of a pair of triangles is constructed.
The legs of one series of reflections lie in a line which is clearly the shortest
distance between two points.
d. In Section 8.17 in addition to all the inversion-related constructions, points
on a line are projected to use the fact that harmonic sets of points are preserved under projection. Also, a triangle connecting the midpoints of the
original is constructed since those points are known to lie on the nine-point
circle, and their positions are well-known relative to the original triangle.
e. In Section 12.1.12, a triangle is constructed so that an earlier theorem
could be applied to a small part of the figure. In addition, a few other lines
required by the previous theorem were constructed.

C.11 Construction Techniques

To do general geometric constructions using a straightedge and compass, you should
have a solid idea of how to do the basic constructions presented in Section 3.5.
Most construction problems are solved by figuring out some properties that the solution has to satisfy, and then to apply some combination of the primitive constructions
referred to above.

C.12. RELABELING

371

It is a bit difficult to divide construction problems into nice categories, but here at
least are a few general approaches that might be useful:
1. Draw a picture of a completed construction and work with it to see what
relations hold. It is often quite easy to draw a solution and work back to the
problem. For example, if you are required to draw a circle tangent to three
lines, draw a circle and then draw three lines tangent to it. From that you
will see that the three lines form a triangle, and then you will have something
to analyze.

2. Sometimes you can use

brute force. If you need to construct a 36 , recall

that cos(36 ) = (1 + 5)/4. Draw an arbitrary segment and

call its length
1. Based on that, construct another segment
of
length
(1
+
5)/4. The right

triangle having legs lengths 1 and (1 + 5)/4 will contain a 36 angle.

3.  If the problem requires the construction of a line or circle tangent to combinations of lines and circles, consider inverting the problem through the
proper circle. See Section C.8.

C.12 Relabeling
Quite often if you wish to prove something for a figure that is completely non-specific,
you can work out the properties for one side or angle, and then just change the labels
in a consistent way to get the result for the others.
As an example, the law of sines states that
b
c
a
=
=
= 2R
sin A
sin B
sin C
for any triangle 4ABC, where R is the radius of the circumscribed circle (See Section A.10.2.4). If you can show that a/(sin A) = 2R you are done, since the triangle
is completely arbitrary, so the labeling does not matter at all. In other words, if you
had taken the same triangle, but had originally exchanged the vertices where you wrote
down A and B, there would be no change in your proof, but your proof would
have shown that the ratio b/sin B is also equal to 2R.

An additional example can be found in Section 4.10 were the existence of the Simson line is shown. The theorem concerns an arbitrary triangle 4ABC and a point M
on the circumcircle. The proof in Section 4.10 depends on the fact that M lies on the
arc CB, but had it been on a different arc, the same argument would hold after we
properly relabeled the vertices of the triangle.

C.13 General Problem Solving Approaches

Here is a list of strategies that can be applied to any mathematical problem, including
geometric ones. The list here is by no means complete, but it may be helpful to review
it if you are stuck on a particular problem.

372

APPENDIX C. GEOMETRIC PROBLEM SOLVING STRATEGIES

Do you understand the problem? Make certain you know what youre looking
for. Name the unknowns. List the known facts.
Did you draw a picture? This is almost always absolutely critical in geometry.
In fact, more than one picture is even better. Try to draw pictures where the result
fails. Often your failed attempts to draw a counter-example make it clear why
the result must be true.
Have you seen or solved a similar problem? If so, what techniques worked
there? Will they work again? If not, why not?
Can you invent a simpler version of the problem that you can solve? If the
problem is for a general polygon, can you do it for a triangle? For a quadrilateral?
If it is about a general triangle, can you solve it for an equilateral triangle? For
an isosceles or right triangle?
Can you decompose the problem? Can you break the problem into parts so
that if you solve all the parts, you have solved the whole problem?
Can you use contradiction? In other words, if you assume that the result you
are trying to prove is false, does that lead to a nonsensical result? If so, you are
done.
Did you use all the information? Sometimes there is extra data in a mathematics problem, but it is rare. If you solve a problem without using some of the data,
it is likely you have made a mistake.
Can you guess, or at least estimate, what the correct answer might be? This
gives a check of your final answer, and may help you to solve the problem.
Can you work backwards? If you assume your final result and work toward
the problem statement, often those steps can be reversed to obtain a solution.
Have you discussed the problem with others? If you can work with others, do
so. Two heads are better than one. Surprisingly, even trying to explain what is
going on to someone totally ignorant of mathematics often helps.
Can you think of different approaches? Maybe vectors will work. Maybe a
symmetry argument. Maybe you can assign coordinates and just grind out the
solution algebraically. Maybe you can use trigonometry.
When you get a solution, can you find another approach? If so, you may have
a way to check your first solution.
Get some sleep. This can be surprisingly effective. You struggle late into the
night on a problem and make no progress, but with a clear head in the morning,
the solution is obvious.

Bibliography

[Abbott, 1952] Abbott, E. A. (1952). Flatland: A Romance of Many Dimensions.

Dover Publications, Inc., New York, seventh edition.
[Altshiller-Court, 1952] Altshiller-Court, N. (1952). College Geometry. Barnes &
Noble, Inc., New York.
[Beman and Smith, 1900] Beman, W. W. and Smith, D. E. (1900). New Plane and
Solid Geomtery. Ginn & Company, Boston.
[Berger, 1987a] Berger, M. (1987a). Geometry I. Springer-Verlag, Berlin, Heidelberg,
New York.
[Berger, 1987b] Berger, M. (1987b). Geometry II. Springer-Verlag, Berlin, Heidelberg, New York.
[Berger et al., 1984] Berger, M., Pansu, P., Berry, J.-P., and Saint-Raymond, X. (1984).
Problems in Geometry. Springer-Verlag, New York, Berlin, Heidelberg, Tokyo.
[Bland, 1819] Bland, M. (1819). Geometrical Problems Deducible from the First Six
Books of Euclid. J. Smith, Cambridge, England.
[Burger, 1965] Burger, D. (1965). Sphereland: A Fantasy About Curved Spaces and
an Expanding Universe. Apollo Editions, New York.
[Conway and Guy, 1996] Conway, J. and Guy, R. (1996). The Book of Numbers.
CopernicusAn Imprint of Springer-Verlag, New York.
[Coxeter, 1989] Coxeter, H. S. M. (1989). Introduction to Geometry. John Wiley &
Sons, Inc., New York, second edition.
[Coxeter and Greitzer, 1967] Coxeter, H. S. M. and Greitzer, S. L. (1967). Geometry
Revisited. The Mathematical Association of America, New York.
[Dudley, 1994] Dudley, U. (1994). The Trisectors. The Mathematical Association of
America, Washington D.C.
373

374

BIBLIOGRAPHY

[Euclid, 1956] Euclid (1956). The Thirteen Books of The Elements. Dover Publications, Inc., New York.
[Eves, 1965] Eves, H. (1965). A Survey of Geometry. Allyn and Bacon, London.

[Godel, 1931] Godel, K. (1931). Uber

formal unentscheidbare satze der principia
mathematica und verwandter systeme, i. Monatsh. Math. u. Phys., 38:173198.
[Greenberg, 1993] Greenberg, M. J. (1993). Euclidean and Non-Euclidean Geometries. W. H. Freeman and Company, New York, third edition.
[Hartshorne, 1997] Hartshorne, R. (1997). Companion to Euclid. Number 9 in Berkeley Mathematics Lecture Notes. American Mathematical Society, Providence.
[Hilbert and Cohn-Vossen, 1983] Hilbert, D. and Cohn-Vossen, S. (1983). Geometry
and the Imagination. Chelsea Publishing Company, New York.
[Johnson, 1929] Johnson, R. A. (1929). Advanced Euclidean Geometry. The Riverside
Press, Cambridge, Mass.
[Johnson, 1917] Johnson, R. A. (Mar. 1917). Directed angles in elementary geometry.
The American Mathematical Monthly, 24:3:101105.
[Kazarinoff, 1961] Kazarinoff, N. D. (1961). Geometric Inequalities. The Mathematical Association of America, Washington D.C.
[Kimberling, 1994] Kimberling, C. (1994). Central points and central lines in the
plane of a triangle. Math. Mag., 67:163187.
[King and Schattschneider, 1997] King, J. and Schattschneider, D., editors (1997).
Geometry Turned On! Dynamic Software in Learning, Teaching, and Research.
MAA Notes. The Mathematical Association of America, Washington, D.C.
[Klarner, 1981] Klarner, D. A. (1981). Mathematical Recreations, A Collection in
Honor of Martin Gardner. Dover Publications, Inc., New York.
[Lawrence, 1972] Lawrence, J. D. (1972). A Catalog of Special Plane Curves. Dover
Publications, Inc., New York.
[Martin, 1991] Martin, G. E. (1991). Geometric Constructions. Springer, New York.
[Meyers, 1996] Meyers, L. F. (1996). Update on william wernicks triangle constructions with three located points. Mathematics Magazine, 69:4649.
[Minsky, 1967] Minsky, M. L. (1967). Computation: Finite and Infinite Machines.
Prentice-Hall, Inc., Englewood Cliffs, N.J.
[Moise, 1963] Moise, E. (1963). Elementary Geometry From An Advanced Standpoint. Addison-Wesley, Reading, Mass.
[Needham, 1997] Needham, T. (1997). Visual Complex Analysis. Oxford University
Press, Oxford.

BIBLIOGRAPHY

375

[Nelson, 1993] Nelson, R. B. (1993). Proofs Without Words: Exercises in Visual

Thinking. The Mathematical Association of America, Washington, D.C.
[Nelson, 2000] Nelson, R. B. (2000). Proofs Without Words II: More Exercises in
Visual Thinking. The Mathematical Association of America, Washington, D.C.
[Ogilvy, 1969] Ogilvy, C. S. (1969). Excursions in Geometry. Dover Publications,
Inc., New York.
[Pedoe, 1988] Pedoe, D. (1988). Geometry, A Comprehensive Course. Dover Publications, Inc., New York.
[Pedoe, 1995] Pedoe, D. (1995). Circles: A Mathematical View. The Mathematical
Association of America, Washington, D.C.
[Polya, 1957] Polya, G. (1957). How to Solve It. Doubleday, second edition.
[Polya, 1962] Polya, G. (1962). Mathematical Discovery: On Understanding, Learning and Teaching Problem Solving. John Wiley & Sons.
[Polya, 1990a] Polya, G. (1990a). Mathematics and Plausible Reasoning. Princeton
University Press, reprint edition.
[Polya, 1990b] Polya, G. (1990b). Mathematics and Plausible Reasoning: Volume II
Patterns of Plausible Inference. Princeton University Press, reprint edition.
[Posamentier and Salkind, 1988] Posamentier, A. S. and Salkind, C. T. (1988). Challenging Problems in Geometry. Dover Publications, Inc., New York.
[Ryser, 1963] Ryser, J. H. (1963). Combinatorial Mathematics. The Mathematical
Association of America, Rahway, N.J.
[Sharygin, 1988] Sharygin, I. F. (1988). Problems in Plane Geometry. Mir Publishers,
Moscow.
[Weeks, 1985] Weeks, J. R. (1985). The Shape of Space: How to Visualize Surfaces
and Three-Dimensional Manifolds. Marcel Dekker, Inc., New York.
[Weisstein, 1999] Weisstein, E. W. (1999). CRC Concise Encyclopedia of Mathematics. CRC Press, Boca Raton.
[Wells, 1991] Wells, D. (1991). The Penguin Dictionary of Curious and Interesting
Geometry. Penguin Books, London.
[Wernick, 1982] Wernick, W. (1982). Triangle constructions with three located points.
Mathematics Magazine, 55:227230.
[Yaglom, 1962a] Yaglom, I. M. (1962a). Geometric Transformations I. The Mathematical Association of America, New York.
[Yaglom, 1962b] Yaglom, I. M. (1962b). Geometric Transformations II. The Mathematical Association of America, New York.

376

BIBLIOGRAPHY

[Yaglom, 1962c] Yaglom, I. M. (1962c). Geometric Transformations III. The Mathematical Association of America, New York.
[Zeitz, 1999] Zeitz, P. (1999). The Art and Craft of Problem Solving. John Wiley &
Sons, Inc., New York.

Index

animations, 38
Apollonian circles, 274
Apollonian gasket, 274
Apollonius problem, 64, 173, 175, 185,
187, 240
arbelos, 184
of Pappus, 182
arc, 324, 362
Archimedes circles, 184
Archimedes midpoint theorem, 288
Archimedes spiral, 133
arcs, 253
area, 298, 366, 368
circle, 38, 324
parallelogram, 319
polygon, 33, 339
trapezoid, 319
triangle, 308, 333
area, maximum, 32
area, of triangle, 308, 333
art, 42
ASA congruence, 306
audience for book, iii
auxiliary lines, 369

AA similarity, 307, 362

AAS congruence, 306
absolute value, 352
accurate drawings, 16
acute angle, 302
acute triangle, 304
Adams circle, 280
adding vectors, 343
Alexanderson, Gerald, iv
algebraic manipulation, 368
algebraic theorems, 368
altitude, 305
analytic geometry, 336
angle, 302
acute, 302
central, 324
complementary, 302
directed, 304
inscribed, 324
obtuse, 302
right, 302
supplementary, 302
triangle sum, 309
trisection, 74
under inversion, 170
vertical, 297, 302, 361
angle bisector, 129, 303, 334, 366
angle calculation, 368
angle trisection, 67, 74
angles, 302
angles between circles, 170
angles, directed, 303
angles, in triangle, 309

Baker, Arthur Latham, i

Benjamin, Arthur, iv
betweenness, 202
Bezier spline, 132
bisect an angle, 51
bisect segment, 50
bisector
angle, 52, 334
book audience, iii
377

378
bound vector, 342, 348
bowtie quadrilateral, 318
Brahmaguptas Formula, 115
Brianchons theorem, 207
dual of, 204
Brocard point, 80
Brocard points, 146
Bush, Ellyn, iv
Cabri Geometry, 1, 15
calculation, 254, 255
cardioid, 34
Cartesian coordinates, 303
caustic curve, 356
Cavalieris principle, 309
center, 323
center of gravity, 304
center of mass, 304
central angle, 324
centroid, 23, 137, 304
centroid, vector form, 350
Cevas theorem, 105
cevian, 106
Chebychev polynomials, 332
chord, 324
Cinderella, 1, 15
circle, 323
area, 38, 324
inversion, 168
nine point, 6, 140
orthogonal, 171
squaring, 74
circle construction, 54
circle inversion, 173
circle theorems, 324
circle, finding center, 54
circle, tangent, 63, 174
circles, 323
mutually tangent, 27
circles, tangent, 62
circumcenter, 23, 46, 137, 306
circumcircle, 306
circumference, 323
circumscribable quadrilateral, 317
circumscribed circle, 54, 55
cissoid of Diocles, 130, 130

INDEX
classical construction, 45, 49, 73
Coffea arabica, iv
collinear, 299
color definition, 245
colors in diagram, 245
combining files, 258
compass construction, 16, 19, 45, 50,
73, 294
compass only construction, 73
complementary angles, 302
complete quadrangle, 233
complete quadrilateral, 233, 233
complex conjugate, 352
complex numbers, 166, 352
computer algebra, 40
computer geometry programs, 1, 2, 15
computer-generated diagrams, 271
concave, 317
concentric circles, 178
conchoid of Nicomedes, 131, 131
concurrence, 363
concurrent lines, 363
concurrent points, 363
concyclic, 299
concyclic points, 3, 6
concyclic quadrilateral, 317
congruence, 294, 361
angles, 361
segments, 361
triangles, 306, 361
congruence, triangles, 306
congruent, 295
conic sections, 130, 202
conic, equation for, 223
conjugate lines, 237
conjugate points, 237
construction, 45, 73
approximate, 74
bisect angle, 52
center of circle, 54
circle from three points, 54
circumscribed circle, 55
classical, 294
copy angle, 51
copy segment, 51
difference of angles, 53

INDEX
division to ratio, 58
external circle tangent, 59
harmonic point sets, 16
heptadecagon, regular, 19
impossible, 74
inscribed circle, 55
internal circle tangent, 59
midpoint, 50
parallel to line, 53
pentagon, regular, 56
perpendicular bisector, 54
perpendicular to line, 52
regular polygons, 76
specific angles, 55
subdivide line, 58
sum of angles, 53
tangent to circle, 57
triangle from sides, 53
construction exercises, 77
construction techniques, 370
construction tools, 47
construction, classical, 49
convex polygon, 321
convex quadrilateral, 317
coordinate geometry, 336
coordinate system, 258
coordinates
Cartesian, 303
change of, 162
homogeneous, 210
in Geometer, 246
polar, 133
projective, 210
copy a segment, 51
copy an angle, 51
copy segment, 51
cos, see cosine
cosecant, 327
cosine, 327
cosines, law of, 334
cot, see cotangent
cotangent, 327
coversine, 327
crackpots, 74
cross product, 211, 220
cross ratio, 235

379
csc, see cosecant
cube doubling, 74
curve envelope, 134
custom colors, 245
cyclic quadrilateral, 317
cyclic quadrilaterals, 96
Davis, Tom, 10
decagon, 319
degenerate conic, 206, 223
degenerate hexagons, 209
Desargues theorem, 193
Descartes circle theorem, 40, 161
descriptive text, 247
diagonal, 233
diagram, 15
diagram appearance, 245
diagram structure, 47
diagram testing, 82
diagrams
computer generated, 34, 35
diagrams, computer generated, 271
diameter, 323
difference of angles, 53
digon, 319
directed angles, 303, 303
directed ratio, 109, 298
Dirichlet domains, 39
disk, 323
displaying numbers, 257
distance formula, 336
dividing segment into ratio, 58
dodecagon, 319
dot product, 347
doubling a cube, 74
Dragon curve, 356
drawing
for publication, 39
drawing manipulation, 21
drawings, accurate, 16
duality, projective, 203
dumbbell curve, 132
ears, polygon, 322
Edit Name, 245
eight point circle theorem, 290

INDEX

380
ellipse, 126, 127, 130, 202
Emerson, Ralph Waldo, 10
encapsulated PostScript, 39
enneadecagon, 319
enneagon, 319
enneakaidecagon, 319
envelope of curve, 134
envelopes, 134
epicycloid, 134, 358
equality, 366
equilateral, 304
equilateral triangle, 362
equivalence relation, 296
escribed circle, 315
Euler line, 281, 281
Eulers formula, 123, 352
excenter, 305
excircle, 305, 315
exercises
construction, 77
proof, 284
exsecant, 327
external tangents, 59
extremes
finding, 32
Fagnanos problem, 32, 280
Fano Plane, 224
Farey circles, 36
Fermat point, 141, 146, 281, 281
Fermat prime, 76
Fermats problem, 176
Feuerbach point, 152
Feuerbachs theorem, 189
Feynman, Richard, 7, 157
figures
file name, 8
misleading, 17
figures, special, 362
finite projective plane, 224
Finsler-Hadwiger theorem, 288
first Napoleon point, 144
floating point numbers, 254
focus of lens, 43
Fourier series, 356
fractal, 271, 356, 357

Fritsch, Rudolf, iv
Gauss line, 287
Gauss, Karl F., 7
geodesic domes, 307
Geometer
area of polygon, 33
how to use, 15
making measurements, 26
proofs, 25
Test Diagram command, 30
Geometer art, 6
Geometer calculation, 255
Geometer CD, 3
Geometer diagram, 47
Geometer files, 48
Geometer proofs, 247
Geometer script, 4, 5, 9, 38, 44, 127,
130, 133, 254, 261
Geometer transformations, 220
Geometers Sketchpad, 1, 15
geometric mean, 57
geometric problem types, 294
geometric transformations, 259
geometry books, 11
geometry websites, 11
Gergonne, 240
Gergonne point, 138
Godel, Kurt, 359
golden ratio, 55, 357
grid drawing, 35
Haeberli, Paul, iv
harmonic numbers, 232
harmonic point sets, 16, 227, 228
harmonic series, 232
harmonic set constructions, 230
harmonically divide, 78
harmony in music, 233
haversine, 327
hendecagon, 319
heptadecagon, 16, 19, 76, 319
heptagon, 319, 357
heptakaidecagon, 319
Herons formula, 113, 256
hexadecagon, 319

INDEX
hexagon, 319
hexakaidecagon, 319
Hilbert curve, 357
homogeneous coordinates, 153, 210
homothetic center, 241
hyperbola, 130, 202
hypotenuse, 304
hypotenuse-leg congruence, 306
hypotrochoid, 131, 131
icosagon, 319
identities, trigonometric, 331
illustrations, 7
imaginary numbers, 352
impossibility proofs, 75
impossible constructions, 74
incenter, 23, 137, 248, 282, 305
incenter-excenter theorem, 290
incircle, 305, 315
index of refraction, 43
inequalities
angles, 366
segments, 366
inequality, 366
infinity
homogeneous coordinates, 211
line at, 196, 211
point at, 71, 162, 165, 172, 195,
197, 207, 229, 367
inscribed angle, 324
inscribed circle, 54, 55, 305, 315
insert file, 258
internal tangent, 59
intersecting secants, 326
intersection, 299
invariants
geometric, 202
projective, 202
inverse
of circle, 168
of line, 166
inversion, 367
complex plane, 166
definition, 164
of points, 162
overview, 162

381
properties, 165
with Geometer, 169
inversion of circle, 173
inversion summary, 172
isogonal conjugates, 148, 290
isosceles trapezoid, 317, 319
isosceles triangle, 304, 362
Japanese temple problem, 283
Johnsons theorem, 291
Jordan curve theorem, 320
jumping points, 253
Kazarinoff, Nicholas, 7
kite, 317
Knuth, Donald, iv
Koch curve, 271, 356
Lamport, Leslie, iv
law of cosines, 334
law of sines, 333
legs, 304
Lemoine point, 152
length, 298
finding unknown, 27
segment, 365
lens focus, 43
limacon, 34
line, 299
inverse, 166
parallel, 365
perpendicular, 365
subdivision, 58
line at infinity, 196
line width, 246
lines, 299
lines, auxiliary, 369
linkage
Peaucellier, 42
lissijous curve, 358
lissijous figure, 131
locus of points, 33
locus problems, 121
manipulation
of drawing, 21

INDEX

382
manipulation, algebraic, 368
Maurer rose, 357
maximizing measurements, 32
maximum area, 32
mean, geometric, 57
measure, 297
angle, 365
segment, 365
measurements, 26
measures, 364
mechanical linkage, 42
medial triangle, 152
median, 124, 304
Menelaus theorem, 108, 363
midpoint, 299
midpoint theorem
Archimedes, 288
midpoint, construction of, 50
Miquels theorem, 21, 30
mittenpunkt, 149
Mohr-Mascheroni theorem, 73
Monges theorem, 24, 197
monogon, 319
Moreton, Henry, iv
Morleys theorem, 22
music, 232
Nagel point, 143
names in diagram, 245
names of primitives, 245
Napoleon points, 144
nephroid, Freeths, 130
nine point center, 140, 140
nine point circle, 6, 140, 140
nine point circle theorem, 25
non-simple quadrilateral, 318
notation, 10, 244
numbers, display, 257
numbers, floating point, 254
obtuse angle, 302
obtuse triangle, 304
octadecagon, 319
octagon, 319
octakaidecagon, 319
orthic triangle, 280

orthocenter, 23, 137, 305

orthogonal circles, 171
Pappus
arbelos, 182
Pappus theorem, 199
dual of, 203
parabola, 33, 130, 131, 131, 202
parallel, 300
parallel line construction, 53
parallel lines, 365
parallel postulate, 310
parallelogram, 316, 319, 363
parametric curve, 127
Pascals theorem, 204
dual, 207
projective version, 206
patterns
finding, 30
Peano curve, 357
Peaucelliers linkage, 42, 188
pedal triangle, 148
pencil of lines, 230
pentadecagon, 319
pentagon, 319
pentagon, regular, 56
pentakaidecagon, 319
perimeter, 320
perimeter bisection, 288
perpendicular, 301
perpendicular at a point, 52
perpendicular bisector, 54, 301
perpendicular lines, 365
point, 299
inversion, 162
on circle, 364
on line, 363
point at infinity, 195
point of intersection, 299
points, 299
points, concyclic, 6
polar, 236
polar circle, 239
polar coordinates, 133
pole, 236
poles and polars, 236, 241

INDEX
Polya, George, 359
polygon, 319
area, 339
convex, 321
ears, 322
interior angles, 323
names, 319
regular, 76, 321
simple, 320
triangulation, 321
polygon area, 33
polygon primitives, 246
polygon theorems, 321
polygon, regular, 76
polygons, general, 319
Poncelets theorem, 67, 123, 264
Pope, Alexander, 10
porism, 179
PostScript, 39
postulate, parallel, 310
power of a point, 118
primitive names, 245
primitive styles, 246
problem solving strategies, 371
problem solving, general, 371
problem types, 294
projection matrix, 221
projective coordinates, 210
projective duality, 203
projective geometry, 193, 195, 229
proof exercises, 284
proof of impossibility, 75
proof, finding, 82
proofs
Geometer, 25
proofs, stepping through, 24
Ptolemys theorem, 6, 111, 175
extension, 176
pythagorean theorem, 311
pythagorean triples, 314
quadrangle, 233
quadrant names, 329
quadrilateral, 316, 316, 319
bowtie, 318
circumscribable, 317

383
complete, 233
concyclic, 317
cyclic, 111, 317, 364
properties, 318
quadrilateral theorems, 318
quadrilaterals, 316
quadrilaterals, cyclic, 96
Renyi, Alfred, iv
radian, 298, 303
radical axis, 116, 118
radical center, 119, 241
radius, 323
ratio, 165, 298, 366
ratio, directed, 109
ratio, dividing segment into, 58
ratio, golden, 55
ratios, 366, 368
ray, 300
re-labeling, 371
reciprocal, 236
rectangle, 316, 363
reduced pythagorean triplet, 314
reflex angle, 303
refraction, index of, 43
regular pentagon construction, 56
regular polygon, 76, 321
relationships, 27
rhodonea, 131, 131
rhombus, 316, 363
right angle, 302, 365
right triangle, 304
rotation, 259, 260, 268, 346, 349
rpn calculations, 254
SAS congruence, 306
SAS similarity, 307, 362
scalene, 304
scaling, 259
script (Geometer), 4, 5, 9, 38, 44, 127,
130, 133, 254, 261
sec, see secant
secant, 324, 326, 327
second Napoleon point, 144
segment, 300
segment length, 365

INDEX

384
segment subdivision, 58
self conjugate, 239
semiperimeter, 315
separation, 202
Shubin, Tatiana, iv, 97
similarity, 296, 362
triangles, 362
similarity, triangles, 306
simple polygon, 320
Simson line, 92, 92
sin, see sine
sine, 327
Snells law, 43
snowflake curve, 271
Soddy circles, 40
Soddy circles, centers, 79
Soddy, Frederick, 162
space-filling curve, 356, 357
special figures, 362
special triangles, 313
specific angles, 55
Spieker center, 152
Spirograph, 43, 358
spline, Bezier, 132
square, 316, 363
squaring a circle, 74
squaring a rectangle, 251
SSS congruence, 306
SSS similarity, 307, 362
Stankova, Zvezdelina, iv, 95, 175
Steiner porism, 179
Steiner-Lehmus theorem, 290
Stewarts theorem, 312
straight angle, 303
straightedge and compass, 16, 19, 45,
50, 73, 294
rules for use, 73
straightedge only construction, 231
string theory, 212
strophoid, 130, 130
style commands, 246
styles, primitives, 246
subdivision of segment, 58
sum of angles, 53
superior triangle, 151
supplementary angles, 302

symmedian, 148, 291

symmetry, 371
tan, see tangent
tangency, 364
tangency, point of, 299
tangent, 299, 300, 323, 324, 327
external, 59
internal, 59
to circle, 325
tangent circle, 6264, 69, 174
tangent line, 323
tangent to circle, 56
tangents, external, 59
Test Diagram command, 30
testing a diagram, 82
testing conjectures, 22
tetradecagon, 319
tetragon, 319
tetrakaidecagon, 319
text in diagrams, 247
Thales theorem, 325
transcendental number, 75
transformations in Geometer, 220
transformations, geometric, 259
translation, 259
transversal, 300
trapezoid, 317, 363
area, 319
isosceles, 317, 319
triangle, 304, 319
angles, 309
area, 308, 333
center, 138
congruence, 306
equilateral, 362
inequality, 177
isosceles, 362
special, 313
spherical, 310
triangle angles, 309
triangle area, 308, 315, 333
triangle center, 23, 137, 154
triangle centers, 138
triangle construction, 53, 124
triangle inequality, 307

INDEX
triangle properties, 307
triangles, 304
triangles, special, 313
triangulation, 321
triangulation of polygons, 321
tridecagon, 319
trigon, 319
trigonometric definitions, 327
trigonometric identities, 331
trigonometric values, 330
trigonometry, 327, 369
Trilinear coordinates, 153
trisect angle, 74
trisection angle, 67
trisection, angle, 74
triskaidecagon, 319
unbound vector, 342
unbound vectors, 342
undecagon, 319
unit circle, 328
unknown length, 27
Varignons theorem, 319
vector centroid, 350
vector notation, 343
vector, bound, 342, 348
vector, unbound, 342
vectors, 342, 369
vectors, adding, 343
vectors, unbound, 342
versine, 327
vertical angles, 302
vertices, 304
volume, 298
Von Aubels theorem, 283
wide lines, 246
witch of Agnesi, 262
zero vector, 343
Zucker, Joshua, iv

385